<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Digital at Virgin Media O2 on Medium]]></title>
        <description><![CDATA[Stories by Digital at Virgin Media O2 on Medium]]></description>
        <link>https://medium.com/@vmo2techteam?source=rss-3a0f72e84a1d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*lgyC7c3m_zhsFoMwVt3ttw.jpeg</url>
            <title>Stories by Digital at Virgin Media O2 on Medium</title>
            <link>https://medium.com/@vmo2techteam?source=rss-3a0f72e84a1d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 04:47:55 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@vmo2techteam/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From Dashboards to Decisions: Meet the Women Building Data Confidence at VMO2]]></title>
            <link>https://medium.com/@vmo2techteam/from-dashboards-to-decisions-meet-the-women-building-data-confidence-at-vmo2-0b07ed69c21f?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/0b07ed69c21f</guid>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[women-in-data]]></category>
            <category><![CDATA[data-democratization]]></category>
            <category><![CDATA[data-culture]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Thu, 05 Mar 2026 12:39:57 GMT</pubDate>
            <atom:updated>2026-03-05T12:39:57.976Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*q8vklKFLrE3jK21uvPXJIw.jpeg" /></figure><p><strong>Data shapes our decisions. Culture decides whether we use it.</strong></p><p>At VMO2, a dedicated team is ensuring data doesn’t just live in dashboards, but drives the decisions made across the company every day.</p><p>This International Women’s Day, we speak to the women helping lead that shift.</p><p>Ahead of their debut on stage at the Women in Data (WID) flagship event, <strong>Louise Le and Lisa Pitherick, Data Evangelists in our Data Democratisation team</strong>, share how they found their way into data, the opportunities that shaped their careers and what it really takes to move organisations from dashboards to decisions.</p><h4><strong>Louise’s Story: From Puzzle-Solver to Culture Shaper</strong></h4><p><strong>Louise, what attracted you to work in data?</strong></p><p>I love a good puzzle. For me you can’t beat the satisfaction of solving a tricky problem.</p><p>Working in data is basically problem-solving day in, day out.</p><p>At university I studied Medical Engineering. Unfortunately (or fortunately!) there’s not much of a market for bionic limbs in the UK, so I had to pivot my career. I was lucky enough to find a data analytics role at a boutique data consultancy and went from there.</p><p><strong>Why are you passionate about building confidence in others, not just delivering insight?</strong></p><p>One thing I learned early in my career is that you can produce the most beautiful dashboards in the world, but there’s no guarantee they’ll be used. In fact, it’s quite likely they’ll collect dust.</p><blockquote>I’ve seen organisations invest millions into technology transformations, and the part that is almost always neglected is enablement.</blockquote><p>What’s the point of doing all the work if no one ends up benefitting from it?</p><p>That’s why I made the decision to pivot from purely technical roles into more strategic ones focused on enablement.</p><p><strong>What is data culture and why is it important?</strong></p><p>A healthy data culture is one where colleagues aren’t afraid of data, but instead embrace it as a tool to make their working lives easier — building the skills and confidence to use it, and bringing it into the decisions they make every day.</p><p>Would you buy a house without considering your financial situation, location or local schools? Would you turn up at the airport without having checked your flight times, or planning your itinerary?</p><p>You’d probably make sure you have the data you need to make an informed decision. It’s the same at work. Without data, you’re left solely with intuition and guesses.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BY7d5Lyz7OcrtbANC_fCgw.jpeg" /><figcaption>Louise and Lisa lead The Martian Frontier data game for colleagues in our Paddington office.</figcaption></figure><p><strong>What does your team do?</strong></p><p>Data Evangelists help colleagues feel more confident using data at work. This means coming up with creative ways to reach people, getting out there to team meetings and offsites, and creating content.</p><p>Our mission is to empower every VMO2 employee to think, act and lead with data and enable everyone to turn insights into value through curiosity, collaboration and action.</p><p><strong>What’s one achievement that you’re particularly proud of, over the last year?</strong></p><p>In 2025, we interacted with over 11,000 colleagues more than 100,000 times. Even more importantly, 94% of colleagues surveyed say our events have improved their confidence using data.</p><p>Martian Frontier, a cooperative data decision-making adventure that I created to help colleagues improve their data literacy and confidence, is one of my proudest achievements to date.</p><p>So far, we’ve delivered to over 800 colleagues. 98% of colleagues have rated it five out of five. It continues to be hugely requested this year, which shows the eagerness of our colleagues to grow their data capability and have fun while doing it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cNcgbstv1IyoQxXYDGCDYQ.jpeg" /><figcaption>Martian Frontier, Louise’s data decision-making game, has been shared with more than 800 colleagues.</figcaption></figure><p><strong>Have there been moments in your career where someone gave you an opportunity that changed your trajectory?</strong></p><p>Definitely. Earlier in my career I worked in consulting, and then as an insights manager in financial services. While there, I was given the opportunity to lead the organisation’s first enterprise-level insights team at just 25 years old.</p><blockquote>Because someone stuck their neck out for me, I was able to seize that opportunity to propel my career.</blockquote><p>It’s not easy to climb the corporate ladder when there are societal injustices that push against you, so I believe it’s vital that those higher up the ladder extend a hand down.</p><h4><strong>Lisa’s Story: From Maths Lover to Data Evangelist</strong></h4><p><strong>Lisa, how did you get into data?</strong></p><p>I grew up in a maths-loving household and it clearly rubbed off on me. I studied maths at university and initially stayed in academia before moving to London to work as a financial modeller.</p><p>When I moved back to Bristol, I stepped into a Data Analyst role and have never looked back. I love helping people solve complex problems by giving them insights they can interpret and act on.</p><p>Since then, I’ve worked in a range of data roles and I’m now a Data Evangelist spreading my love of data.</p><p><strong>What do you do, day to day?</strong></p><blockquote>An evangelist is someone who passionately spreads a message, and that’s exactly what we do for data.</blockquote><p>As Data Evangelists, we break down the fear that can come with it, so that people feel empowered to use it.</p><p>Day-to-day we run events and interactive sessions, share engaging content, spotlight great work and find creative ways to build confidence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_Q0q5S4F2PxcGU8Rh4CENg.jpeg" /><figcaption>Lisa enjoys sharing her love of data with colleagues.</figcaption></figure><p><strong>Have there been moments in your career where someone gave you an opportunity that changed your trajectory?</strong></p><p>I wouldn’t point to one single opportunity, but I’ve been incredibly fortunate to have managers who really believed in me. They backed me and trusted me, which really helped build my confidence and belief in myself.</p><p>Sometimes having someone champion you can make all the difference.</p><p><strong>This year’s IWD theme is “Give to Gain”. What does that mean to you?</strong></p><p>To me, “Give to Gain” is about sharing knowledge, confidence and opportunities.</p><blockquote>When we give colleagues, especially women, access to tools, skills and a safe space to learn and ask questions, we gain stronger voices, better ideas and more diverse perspectives in our decision-making.</blockquote><p><strong>If we get this right, what will be different about how decisions are made at VMO2?</strong></p><p>If we truly embed data into how we work, people will see it as something that supports them, not something scary or complicated.</p><p>Colleagues will be able to easily access the information they need, trust it and feel confident using it in their day‑to‑day work. Teams will talk about data in simple, consistent ways, making collaboration smoother.</p><p>Most importantly, people will feel comfortable testing ideas and learning from the results, so the whole organisation keeps getting better.</p><p><strong>What advice would you give to women considering a career in data today?</strong></p><p>Be curious and ask questions. Earlier in my career, it took me a long time to build up the confidence to ask questions, especially when I was the only woman on the team. I worried about looking stupid. But once I got over that fear, I learnt so much from my colleagues, and everyone was always supportive.</p><p>There has never been a more exciting time to work in data, so don’t wait.</p><p>Want to come work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace/SearchJobs/?jobRecordsPerPage=8&amp;jobOffset=8">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0b07ed69c21f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Customer trust isn’t a given. Here’s how we earn it.]]></title>
            <link>https://medium.com/@vmo2techteam/customer-trust-isnt-a-given-here-s-how-we-earn-it-62de4e1c0765?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/62de4e1c0765</guid>
            <category><![CDATA[data-privacy]]></category>
            <category><![CDATA[data-democratization]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Wed, 28 Jan 2026 10:08:37 GMT</pubDate>
            <atom:updated>2026-01-28T10:08:37.318Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*GStkNU12zNrOrKWvwZZUiw.jpeg" /></figure><h4><strong>Customer trust takes years to build and seconds to lose.</strong></h4><p>Every decision we make with data either strengthens that trust or puts it at risk.</p><p>Data privacy is often misunderstood as a barrier, but in reality, it’s one of the strongest enablers of customer trust.</p><p>For <strong>Data Privacy Day, </strong>we sat down with <strong>James Boylett, Head of Privacy Management in our Data Democratisation team at VMO2</strong>, who shares how we earn customers’ trust every day and why doing the right thing with data is everyone’s business.</p><p><strong>James, why does data privacy matter so much right now?</strong></p><blockquote><strong><em>Trust is incredibly fragile. It takes a long time to build and can be broken in a moment. If we compromise privacy, regaining that trust is incredibly hard.</em></strong></blockquote><p>We’ve seen high-profile cases, ongoing media attention and now the rapid growth of AI. People understand more than ever that their data has value, and that it can have real consequences.</p><p>We’re a data-rich company, and we should feel honoured that customers trust us with so much information. Protecting that trust is essential to our ability to do business.</p><p><strong>You’re our Head of Privacy Management. What does your team do?</strong></p><p>I lead the Privacy Management function, and we exist to support the entire organisation to use data confidently and responsibly.</p><p>Most of our time is spent partnering with teams — particularly on new products and programmes — helping them understand their privacy risks and design the right controls.</p><p>We supported over 250 projects across the company last year, all with different challenges.</p><p>Privacy is often seen as a blocker, but that’s simply not the case.</p><blockquote><strong><em>Our role isn’t to say no. It’s to help the business achieve the right outcomes for customers. Finding ways to do things in a customer-centric way, while still enabling progress, is what motivates me.</em></strong></blockquote><p><strong>What drew you to data privacy and what do you love about your job?</strong></p><p>My first proper job was in local government, working in IT risk management.</p><p>What really drew me to privacy was seeing the very real, human impact it has on people’s lives — especially in areas like social care. Decisions about whether to share data with partner agencies like the NHS or the police could genuinely change outcomes for individuals.</p><p>We hold huge amounts of data at VMO2, and with that comes power. Used well, it can deliver amazing outcomes for customers. Used badly, it can cause serious harm. It’s this responsibility that keeps me interested in privacy.</p><p><strong>What does data privacy really mean for customers and colleagues?</strong></p><p>Privacy is fundamental to customer trust.</p><p>If customers don’t trust us, they won’t want to do business with us. If they do trust us, they’re more confident allowing us to use their data, which enables better services and decisions.</p><blockquote><strong><em>Giving customers the right level of control, clarity and confidence that we care about their privacy builds that trust, and it should be seen as a business differentiator rather than a restriction.</em></strong></blockquote><p>For colleagues, trust means confidence — knowing what you can and can’t do with data, understanding the risks, and being supported to use data responsibly to achieve the right outcomes.</p><p><strong>What’s the biggest privacy challenge organisations face today?</strong></p><p>Balancing speed and safety. Businesses need to move quickly, adapt to market changes and meet customer expectations — without diluting standards or protections.</p><p>Traditional approaches to privacy don’t scale. Long assessments for every small change or relying on a single central team doesn’t work.</p><p>We need to empower individuals to make informed decisions, supported by subject-matter experts for the more complex issues.</p><p><strong>What’s your proudest moment at VMO2 so far?</strong></p><p>I’m particularly proud of our support for the 3G shutdown. The original design planned to use network location data to identify customers still relying on 3G for highly targeted notifications.</p><p>Because location data is highly sensitive, we worked with the delivery team to rethink the approach. We designed an approach that used less intrusive data like device type and service address to reach the right customers and put extra protections in place for vulnerable users.</p><p>The result was the same outcome, with less intrusion and more trust.</p><p>It’s a pattern we use often: start with the purpose, choose the least intrusive effective data, and bake the guardrails in from day one.</p><p><strong>Can you share an example of a project that moved more smoothly because privacy was considered early?</strong></p><p>Verity, Data Democratisation’s Analytics AI platform, is a great example. We brought privacy in at the design stage, not at sign-off, and worked closely with our teams to map responsible AI risks up front.</p><p>We agreed clear guardrails early, limited the data the tool could access and made sure sensitive personal data wasn’t in scope from day one.</p><p>Because those decisions were made up front, the project moved faster, avoided late rework and is now a strong baseline for future AI projects.</p><p><strong>How has AI changed the game for data privacy?</strong></p><p>AI is a huge game changer and the ability to process data faster and draw new inferences increases dramatically. This creates incredible opportunities to better support customers, but also greater responsibility.</p><p>We need to be thoughtful, responsible and deliberate in how we use data, and avoid taking unnecessary risks.</p><p>Our focus is building strong guardrails, educating colleagues and making sure privacy considerations are built in early — so we’re not reacting later to changes in law or policy.</p><p><strong>What role do our colleagues play in helping us protect customer trust?</strong></p><blockquote><strong><em>Data Privacy is everyone’s responsibility. That’s why we’re focused on helping colleagues understand how privacy shapes design and decision-making and giving them the frameworks and support to take ownership.</em></strong></blockquote><p>We’re increasingly seeing teams ask good questions early, like do we really need this data? Can it be minimised or anonymised?</p><p>Every decision we make with data either strengthens customer trust or puts it at risk. When we get it right, customers feel it, the business benefits and trust grows.</p><p>Want to work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62de4e1c0765" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[National Inclusion Week: Data Doesn’t Lie? Think Again.]]></title>
            <link>https://medium.com/@vmo2techteam/national-inclusion-week-data-doesnt-lie-think-again-ea98fd348d6a?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/ea98fd348d6a</guid>
            <category><![CDATA[inclusion]]></category>
            <category><![CDATA[diversity-in-tech]]></category>
            <category><![CDATA[diversity-and-inclusion]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Wed, 17 Sep 2025 12:29:02 GMT</pubDate>
            <atom:updated>2025-09-17T12:29:02.511Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rDqt7IF3tCgWWH8E4NsQaA.jpeg" /><figcaption>Jolene Marshall, Data Governance Manager in our Data Democratisation team at VMO2</figcaption></figure><p>Data drives more decisions than ever, but what if the numbers are stacked against you? From recruitment algorithms that exclude women, to credit systems that undervalue female applicants, one thing is clear: without inclusion, AI and data risk reinforcing the very biases organisations have worked to overcome for decades.</p><p>For National Inclusion Week, we sat down with Jolene Marshall from our Data Governance team at VMO2, to talk about how we redress the balance and why it’s time to treat inclusion in data as a priority, rather than an afterthought.</p><p><strong>Jolene, why should inclusion in data and AI matter to everyone?</strong></p><p>Because when we get it wrong, we risk building systems on assumptions and stereotypes instead of reality.</p><p>We’ve seen examples where AI recruitment tools rejected female candidates because the training data reflected male-dominated leadership teams, or credit models gave men far more credit than women with the same risk profile. Even facial and voice recognition systems are skewed toward white people and can’t reliably pick up certain accents, like Scottish.</p><p>Too often the path is pre-set: you decide the goal, collect the data to match, and never stop to ask if you’re framing the problem in the right way.</p><p>That’s why inclusion matters. It’s about representation and perspective: who’s in the room when we’re defining the problem? What questions are we asking, and which groups might we be overlooking?</p><p><strong>What happens if organisations don’t take this seriously?</strong></p><p>Systems can hard-code bias. In one banking system I worked with, selecting ‘Doctor’ as a title automatically defaulted the gender to male. You couldn’t change it. Mandatory Mr/Mrs/Miss fields also reinforce outdated ideas about identity.</p><p>These may feel like small details, but together they build systems that exclude, frustrate, or even harm people. If we rely on these blunt categories, we risk building services on stereotypes rather than reality.</p><p>It’s also personal. In data roles, there’s still a heavy bias toward candidates with technical backgrounds like computer science degrees. I don’t have one. I started as a lawyer, and I know my CV would be filtered out by AI recruitment tools because of that. Yet my experience proves you don’t need to be a coder to work in data. If we only recruit from one narrow pipeline, we miss out on other perspectives.</p><p>It’s critical to recognise that data isn’t automatically factual. It reflects the questions we ask and the way we frame them. If we treat data as truth, we just reinforce historic inequalities.</p><p><strong>Beyond the risks, where do you see the biggest potential?</strong></p><p>The potential for good. In healthcare, for example, AI could process huge data sets and highlight patterns in women’s health that doctors might miss or undervalue — helping to challenge biases in medical treatment and improve outcomes.</p><p>But it comes with responsibility. We need to avoid over-reliance, environmental costs, and the “AI for AI’s sake” trend. The exciting part is using AI where it really matters, to redress inequalities and improve lives.</p><p><strong>What difference can inclusive data make for our customers and business?</strong></p><p>At VMO2 we serve really varied groups, and if you look closely there are patterns of behaviour we can learn from. If we only use data to push more products, we’re missing the point.</p><p>Inclusive data lets us design services that genuinely meet people’s needs, rather than just pushing more products. When customers see that benefit, they’re more willing to share their data, which builds trust and leads to better business outcomes.</p><p><strong>What practical steps can organisations take to make data more inclusive?</strong></p><p>First, make data accessible. Not everyone is comfortable with dashboards. We need to demystify it and provide different ways of engaging with it — written, visual and interactive — and show that data literacy is for everyone. Second, bring in external perspectives. Otherwise, teams risk groupthink. Workshops, guest speakers or hackathons can challenge assumptions and spark new ways of thinking.</p><p>And finally, always tie data back to real business questions and customer outcomes. Data for its own sake has no value. It has to answer something meaningful.</p><p><strong>How has your own background influenced how you approach inclusion?</strong></p><p>My career began in law, which taught me to question everything. Not just whether we <em>could</em> do something with data, but whether we <em>should</em>. Later, in financial regulation, I saw how some organisations gamed the system for profit at consumers’ expense.</p><p>That reinforced my interest in ethics and fairness.</p><p>Personally, being a woman and neurodivergent has shaped how I approach inclusion. I know how challenging environments can be when they’re designed for only one type of person.</p><p>That’s why I adapt interviews to support neurodivergent candidates and bring junior colleagues into senior meetings to give them exposure. For me, inclusion isn’t about favouring one group. It’s about removing unnecessary barriers.</p><p><strong>What’s your advice for anyone who wants to think more inclusively about data?</strong></p><p>Stay curious. Pause and ask: am I just following assumptions? What perspectives might I be missing? Go to conferences, read widely, and get outside your bubble.</p><p>Data is a tool, not the answer. It won’t solve problems on its own. We need to bring human judgement, diverse perspectives and good judgement to get the outcomes we want.</p><p>Want to work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ea98fd348d6a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Governance, but without the Red Tape: The VMO2 Way]]></title>
            <link>https://medium.com/@vmo2techteam/governance-but-without-the-red-tape-the-vmo2-way-a265ac3d99a5?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/a265ac3d99a5</guid>
            <category><![CDATA[data-transformation]]></category>
            <category><![CDATA[data-democratization]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[trust]]></category>
            <category><![CDATA[data-governance]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Fri, 20 Jun 2025 09:57:52 GMT</pubDate>
            <atom:updated>2026-11-05T12:39:26.278Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZEp_uRlGCykf5QfjNGcoaw.jpeg" /><figcaption>She used to be a volunteer police officer. Now she’s proving governance is about empowerment, not control.</figcaption></figure><blockquote><strong><em>“People think we’re the data police. We’re not. We’re here to help.” — Helen Gardiner, Head of Data Governance at VMO2</em></strong></blockquote><p>One of the unlikely heroes of our digital transformation? Data Governance.</p><p>Too often it’s seen as a blocker. But at VMO2, we’re changing that.</p><p>Shedding its reputation as a red-tape function, data governance is helping us create trust and put high-quality data into the hands of every colleague.</p><p>At the heart of this change is our Head of Data Governance, Helen Gardiner. We catch up with her to find out how data governance can support innovation and why trust — not control — is the secret ingredient in any data-driven transformation.</p><p><strong>Helen, why should everyone care about data governance?</strong></p><p>We’re on a journey to make data accessible across the entire organisation, but that only works if people trust our data.</p><p>Data governance is often mistaken for being the data police, but it’s actually the opposite. It’s about empowerment. It gives people the confidence to use data by making it trustworthy, consistent and well-managed.</p><p>It means putting the right standards, tools and processes in place to ensure data is reliable, used responsibly, and fit for purpose. It reduces risk, improves quality, and enables better, faster decisions.</p><blockquote><strong><em>“If we want to be a truly data-driven organisation, people need to believe in the data they’re using. That belief starts with governance.”</em></strong></blockquote><p><strong>What’s a popular misconception about data governance?</strong></p><p>The biggest misconception we deal with is that governance is about policing data and telling people what they can’t do. It’s really the opposite. We’re here to enable people.</p><p>One way we do this is through our data governance clinics. These are open sessions where colleagues can bring questions, challenges, or pain points</p><blockquote><strong><em>“People see we’re not there to control, but help. We’re working with teams to embed governance in a way that supports, rather than stops, progress.”</em></strong></blockquote><p><strong>What does your role as Head of Data Governance involve?</strong></p><p>I lead the team that sets the standards for how we govern data at VMO2, and I ensure those standards are actually lived across the business.</p><p>Our focus is on delivering real value: making data more trusted, consistent and usable. We do that through tools like our business glossary and data catalogue, which ensure everyone’s working from the same definitions.</p><p>We also surface data health, so people can see where issues exist and what we’re doing to fix them. We track adoption, measure maturity — and always come back to our core aim: building trust in the data. Because if people don’t trust it, they won’t use it.</p><p><strong>What are the top priorities for you and your team right now?</strong></p><p>We’re aiming to create a best-in-class data discovery experience so our colleagues can easily search, find, and self-serve trusted data and dashboards. This needs to be intuitive, with clear results and business-friendly language. To do this, we’re focused on getting the most out of our data catalogue and marketplace tools.</p><p>One challenge our team has faced is that sometimes our colleagues spot data issues before we do, which undermines trust. That’s why we’re also working to detect and address data quality issues proactively, using a data observability tool.</p><p><strong>How do you balance control with innovation and access to data?</strong></p><blockquote><strong><em>“We start from a principle of openness — everyone can access everything — and then scale back only where absolutely necessary, such as with personal or commercially sensitive data.”</em></strong></blockquote><p>This approach shifts the mindset from control to enablement. That said, there are areas where control is essential, like our business glossary. Consistency in definitions is non-negotiable because it underpins trust.</p><p>If people find conflicting definitions for the same metric, trust erodes quickly. So we’re happy to maintain control where it protects the integrity of the data.</p><p><strong>What would you be doing if you weren’t in data?</strong></p><p>I’ve always been a bit envious of civil engineers. It must be so satisfying to look at a bridge or a tunnel and say, “I built that.” I’d probably have enjoyed being one, but I wasn’t good enough at maths at school.</p><p>Another option — maybe a bit of a cliché for someone with a geography degree — but I absolutely adore an Ordnance Survey map. I’d have loved to work for them be a ranger in a national park. I just really love being outdoors in the hills and mountains.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OCYAvyXJwp9N0UkB_RIjxA.jpeg" /><figcaption>When not at work, Helen enjoys hiking with her dog, Connie.</figcaption></figure><p><strong>What’s one thing most people don’t know about you?</strong></p><p>I used to be a volunteer police officer, a Special Constable. I’d work a full day and then go on shift from 7 pm to 3 am. It gave me experiences you’d never get in any other job. It taught me a lot about resilience and perspective.</p><p><strong>What keeps you up at night?</strong></p><p>Rotator cuff tendinopathy in my left shoulder! If you’ve had it, you know the pain is real. I’ve just had a steroid injection, so I’m hoping that helps me finally get some sleep.</p><p><strong>What book should every data leader read?</strong></p><p><em>Invisible Women</em> by Caroline Criado Perez. It exposes the gender bias in data collection and use and shows the very real impact of that bias.</p><p>Want to work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a265ac3d99a5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data in Action: Powering VMO2’s Fight Against Fraud]]></title>
            <link>https://medium.com/@vmo2techteam/data-in-action-powering-vmo2s-fight-against-fraud-4c9d37d0f8d1?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/4c9d37d0f8d1</guid>
            <category><![CDATA[transformation]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[fraud-detection]]></category>
            <category><![CDATA[data-driven]]></category>
            <category><![CDATA[fraud-prevention]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Thu, 22 May 2025 07:26:17 GMT</pubDate>
            <atom:updated>2025-05-22T07:26:17.103Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*FWo4x9zl1rEta-kA6xTIYg.jpeg" /><figcaption>Murray Mackenzie, VMO2 Director of Fraud, and Mauro Flores, EVP of Data Democratisation at VMO2</figcaption></figure><p>When Murray Mackenzie joined Virgin Media O2 as Director of Fraud, his mission was clear: transform how the business tackles fraud and protects customers.</p><p>Since then, we’ve prevented <strong>£250M+ in fraud</strong> and blocked <strong>168M malicious texts</strong> thanks to our commitment to <strong>make data accessible to all 16K+ colleagues </strong>through our Data Democratisation programme.</p><blockquote><strong><em>“The goal was to reduce our exposure and protect our customers. And from the start, data was at the heart of it.” — Murray Mackenzie</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*16tsp93hrHBbcIO-VXkvPQ.jpeg" /><figcaption>Murray is driving a data-first approach to prevent fraud and protect our 23M+ customers.</figcaption></figure><p>Fraud was seen as a priority area, and Murray’s team was one of the first to adopt Tableau, our all-company analytics platform, with dedicated support from our Data Democratisation team.</p><p>“It’s not just about dashboards,” says Murray. “It’s about understanding the data, prioritising it and acting fast. That’s where the value really comes in.”</p><p>The result? Millions of pounds worth of fraud stopped in its tracks, and a cultural shift that’s empowering people across VMO2 to solve fraud before it even hits Murray’s desk.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*ZrPft0tPVtDZYF3MoVrcWQ.jpeg" /><figcaption>With support from our Data Democratisation team, Murray’s team has put data to work against fraud.</figcaption></figure><h3><strong>Fraud’s fast, but data’s faster</strong></h3><p>“Our goal is to make data accessible, trusted and useful for every colleague — so they can move faster, make smarter decisions and drive impact,” says Mauro Flores, EVP of Data Democratisation at VMO2.</p><p>And that’s exactly what happened earlier this year, when the fraud team spotted something unusual in the data — a pattern that didn’t match typical customer behaviour.</p><p>Within 40 hours, a new fraud attack had been identified. Within 72 hours, the specific journey the fraudsters were exploiting was understood and disabled. And just two days later, it was rebuilt, protected and back online.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*0cPWkdVvowgD1LOOJd0MRQ.jpeg" /><figcaption>Mauro joined VMO2 with a mission: to democratise data and empower every colleague to use it confidently.</figcaption></figure><p><strong>That single insight prevented over £1 million in fraud.</strong></p><p>One of the most striking shifts Murray has seen isn’t in tools or technology, but in ownership, with colleagues across the business treating data not as someone else’s responsibility, but part of their day job.</p><h3><strong>Making data everyone’s business</strong></h3><p>“People across the business are coming to me and saying not just, ‘Here’s a problem,’ but ‘I’ve already solved it,’” says Murray.</p><p>A powerful example is our Commercial team, who used their own dashboard to identify voucher abuse, analyse the issue, and implement a fix — all before alerting the fraud team.</p><p>And they’re not alone. The team believes that about half of the insights that now drive fraud prevention at VMO2 don’t originate from the fraud team at all — they come from colleagues across the business, empowered by access to data.</p><h3><strong>Building our data culture</strong></h3><blockquote><strong><em>“There’s a culture change that needs to happen when it comes to working with data. Part of our role is to reduce the fear factor by providing context, capability and confidence.” — Mauro Flores, EVP, Data Democratisation</em></strong></blockquote><p>This shift has been supported by leaders like Murray, who don’t just champion the use of data, but actively use it themselves and lead by example.</p><p>“I was in a senior meeting recently where someone said fraud was up 28%. I was able to check the live dashboard there and then. I said, ‘It’s not, and here’s the real story.’ That’s a game-changer,” says Murray.</p><p>This ability to respond immediately has transformed how fraud is discussed across the business.</p><p>Real-time alerts through Tableau Pulse now notify Murray, senior leaders, and our Executive Committee as risks emerge.</p><p><strong>From reactive to predictive fraud prevention</strong></p><p>“I don’t have to go looking for problems anymore. Pulse surfaces them automatically. I can even ask it to create a view for me,” says Murray.</p><p>Fraud data is also challenging long-held assumptions, especially about who’s at risk.</p><p>“There’s a myth that it’s just the elderly,” explains Murray. “But the data tells us it’s everyone. Anyone with a device can be targeted.”</p><p>He adds: “Helping that education drive has been absolutely essential.”</p><p><strong>What’s next in our fight against fraud?</strong></p><p>Murray’s long-term ambition is for data and insight to be shared across companies and industries, creating collective intelligence too powerful for criminals to outrun.</p><p>“It’s a long road,” he says. “There’s regulation and privacy to navigate. But that’s the biggest, shiniest opportunity we’ve got.”</p><p><strong>How can you Power Up with Data?</strong></p><p>For anyone who wants to get started with data but feels hesitant, Murray has a simple message.</p><p>“There’s a myth that data is complicated. It’s not. The tools we have now, like Tableau, are incredibly easy to get started with.”</p><p>“Don’t be afraid of the rabbit hole. Sometimes you’ve just got to dive in, head-first, and see what you can find.”</p><p>It’s a mindset that captures the spirit of Data Democratisation at VMO2: that great decisions come from great data, and that great data belongs in everyone’s hands.</p><p>Want to come work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4c9d37d0f8d1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Resetting Bias in Data & AI: Lessons from Women in Data 2025]]></title>
            <link>https://medium.com/@vmo2techteam/resetting-bias-in-data-ai-lessons-from-women-in-data-2025-07ba7f193fd0?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/07ba7f193fd0</guid>
            <category><![CDATA[inclusive-ai]]></category>
            <category><![CDATA[women-in-data]]></category>
            <category><![CDATA[diversity-and-inclusion]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[inclusive-data]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Mon, 31 Mar 2025 10:30:33 GMT</pubDate>
            <atom:updated>2025-03-31T10:30:33.187Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Lj0_iDyQ1_rhHvmWNg9XwA.jpeg" /><figcaption>Our Data Democratisation and DPPA colleagues on stage at the Women in Data flagship event.</figcaption></figure><p>We were proud to represent Virgin Media O2 at this year’s <strong>Women in Data flagship event</strong>, with a panel discussion titled <strong>‘CTRL + ALT + Include: Resetting Bias in Data and AI’</strong>.</p><p>Bringing together experts from across our Data Democratisation and Data Platforms, Products &amp; AI (DPPA) teams, the panel sparked an honest conversation about bias in data and AI, and how we all have a role in changing it.</p><h3><strong>Designing for inclusion from the start</strong></h3><p>Our panel explored what it takes to design for inclusion from the ground up — from recruitment pipelines to platform design, to decision-making.</p><p>Bias in data and AI isn’t just a technical problem, it’s a human one, says Jolene Marshall, Data Governance Manager in our Data Democratisation team.</p><p>It starts with how we define the problem, who’s in the room to solve it, and what systems we’re working with.</p><p>“Data holds a lot of power — but it all comes down to human design. We have to ask the right questions to make sure the outcomes are fair and unbiased,” said Jolene.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mfSiT0epvIGrVaA4avaSmg.jpeg" /><figcaption>Jolene Marshall, Data Governance Manager, and Shahd Girsi, Lead Machine Learning Engineer</figcaption></figure><p>Jolene reminded us that inclusion begins with the intent and design behind every product, dataset, and model.</p><p>This means asking: Who gets to frame the problem? Who designs the model? And who’s held accountable when things go wrong?</p><p>In other words — inclusion doesn’t happen by accident. It needs to be built in from the beginning.</p><h3><strong>Inclusive hiring and career growth</strong></h3><p>Inclusion goes beyond platforms or algorithms — it extends to the structures, processes, and mindsets that shape how we build teams, make decisions, and create opportunities — especially in areas like hiring.</p><p>“We hire based on skills and potential — not just years of experience,” said Sean Andrew, Head of Data Go To Market and Services and Support, in our Data Democratisation team.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FGQak9uBKKkUyrT9JgHIow.jpeg" /><figcaption>Shahd Girsi, Lead Machine Learning Engineer, and Sean Andrew, Head of GTM and Services and Support</figcaption></figure><p>Sean shared how VMO2 is using AI to write more inclusive job descriptions, structure interviews around competencies, and support fairer hiring decisions with diverse interview panels.</p><p>He also reflected on unconscious bias training as a powerful learning moment and a reminder that inclusive hiring depends not only on process, but on self-awareness and mindset.</p><p>“I used to think I was pretty self-aware — but unconscious bias training really opened my eyes. It challenged me in ways I didn’t expect, and it’s something I’d recommend to everyone,” Sean said.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CC-zgVvBeN-Oyh0wQTM3iQ.jpeg" /><figcaption>EVP Data Democratisation, Mauro Flores, and Bhaga Reddy, Director of Data Empowerment and Collaboration</figcaption></figure><p>Our Executive Director of Data Empowerment and Collaboration, Bhagya Reddy, built on this by highlighting the importance of sponsorship — especially for women returning to work or navigating career growth.</p><p>“Find a sponsor, not just a mentor,” said Bhagya. “A mentor offers guidance. A sponsor creates opportunities.</p><p>“Find someone who will back you and push you forward.”</p><p>For female colleagues returning to work following maternity leave or a career break, this active sponsorship can make all the difference.</p><h3><strong>Allyship and shared accountability</strong></h3><p>None of this work can sit solely on the shoulders of women. That’s why our panel also stressed the importance of allyship and shared accountability.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y0_hzS8sfM0weJXd8p9VKg.jpeg" /><figcaption>Shahd Girsi, Lead Machine Learning Engineer in our DPPA team</figcaption></figure><p>“We need male allies at the table — not just as guests, but active participants in driving change,” said Shahd Girsi, Lead Machine Learning Engineer in our DPPA team.</p><p>“No matter how far along you are, you still need people to help shape your vision.”</p><p>Shahd’s message was clear: inclusion isn’t just a women’s issue — it’s a leadership issue. Male allies have a role to play in shifting not just policies, but workplace cultures.</p><h3><strong>What’s next?</strong></h3><p>This panel was just one conversation. The work continues.</p><p>In data at Virgin Media O2, we’re committed to inclusive hiring powered by AI and human insight, building ethical and transparent design practices into our AI products, supporting returners and future talent through mentoring and outreach and educating ourselves on unconscious bias.</p><p>This year, we reached 41% female representation in our Data Democratisation team. This isn’t the finish line, but it demonstrates progress is happening.</p><p>Bias won’t be reset by algorithms alone. It takes intentional culture change, collective accountability, and everyday acts of inclusion.</p><p>Until then we will keep pushing for a data future that is inclusive by design.</p><p>Want to come work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace/SearchJobs/?jobRecordsPerPage=8&amp;jobOffset=8">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=07ba7f193fd0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Unlocking the Power of Neurodiversity in Data]]></title>
            <link>https://medium.com/@vmo2techteam/unlocking-the-power-of-neurodiversity-in-data-5045f89b593f?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/5045f89b593f</guid>
            <category><![CDATA[neurodiversity]]></category>
            <category><![CDATA[data-democratization]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Wed, 19 Mar 2025 14:40:44 GMT</pubDate>
            <atom:updated>2025-03-19T15:03:30.196Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*KE4mP6XuwGzzAcAEAA_Xlg.jpeg" /><figcaption>Karl Nicholson and Harriet Chapman are in our Data Democratisation team at VMO2.</figcaption></figure><p>What if the way you think, the very thing that sets you apart, is actually your greatest strength? What if it is a competitive advantage in the workplace?</p><p>In the world of data, where problem-solving, pattern recognition, and innovation are key, neurodivergent colleagues are discovering that not only are their differences their greatest superpower, but that organisations are starting to take note.</p><p>For <strong>Neurodiversity Celebration Week</strong>, we caught up with <strong>Harriet Chapman, a Digital Graduate at VMO2, and Karl Nicholson, Data Culture Lead in our Data Democratisation team</strong>, about their journeys, the challenges they have faced and why embracing neurodivergence in data is not just valuable — it is essential.</p><h3><strong>Harriet’s Story: Thriving in Data Through Neurodivergence</strong></h3><p><strong>Harriet, how has neurodiversity shaped your career in data?</strong></p><p>I have always been drawn to problem-solving, numbers, and patterns. These things come naturally to me because of my neurodivergence. I studied computer science at university and deliberately chose a course where neurodivergence was understood. When I learned that most of my classmates were also neurodivergent, I knew I was in the right place.</p><p>Now, I am on the Graduate Scheme at VMO2, working with the Data Democratisation team. So far, I have worked in forecasting, planning, and data modelling. It is a field where different ways of thinking are not just accepted; they are needed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FCLz8iGtYiGUAA1kTCzm7w.jpeg" /><figcaption>Harriet Chapman is currently doing a rotation in data through VMO2’s Digital Graduate scheme.</figcaption></figure><p><strong>What strengths do neurodivergent individuals bring to data?</strong></p><p>We see things differently. We recognise patterns others might miss, approach problems in unique ways, and aren’t afraid to challenge traditional thinking. That’s why so many neurodivergent individuals excel in technical fields.</p><p><strong>Can you share a bit about your neurodivergence and how it has shaped you?</strong></p><p>My mother suspected I was neurodivergent from a young age, but teachers dismissed her concerns, saying I fit within “average expectations.”</p><p>It wasn’t until a tutor reviewed my English work that my parents sought a formal diagnosis. I was diagnosed with dyslexia at 15 and dyspraxia at 18.</p><blockquote>That moment was a relief. It explained why I struggled with tying shoelaces, catching a ball, or processing written information as quickly as others.</blockquote><p>It also helped me understand why my written work never matched my verbal ability. Every school report would say, <em>“When Harriet’s written work matches her verbal ability, she will be an A student.”</em></p><p>But I did not know how to bridge that gap. My diagnosis gave me clarity and access to support which made a huge difference.</p><p><strong>What challenges have you faced at work?</strong></p><p>Ambiguous expectations can be frustrating. Phrases like “Do this if you can” leave me wondering do they actually need me to do it, or is it optional?</p><p>I also struggle with working memory. If someone lists ten tasks for me to do, I’ll probably only remember the first two. But I’ve found that if I explain how my brain works, most colleagues are very supportive.</p><p><strong>How can organisations better support neurodivergent employees?</strong></p><p>Clear expectations and advance notice for meetings are incredibly helpful. Leaders should also recognise that not everyone processes information at the same speed, so allowing time before expecting a response makes a big difference.</p><blockquote>More training and guidance at all levels — not just for leaders — would help. We work with our peers daily, and if everyone had a better understanding of how different brains work, we’d be able to communicate more effectively.</blockquote><p><strong>What’s your advice for neurodivergent individuals considering a career in data?</strong></p><p>Go for it! Neurodivergent minds bring valuable perspectives, especially when it comes to problem-solving. Data is all about finding insights, and different ways of thinking help uncover things others might miss.</p><p><strong>Why is it important to have neurodivergent voices in data?</strong></p><p>Neurodivergent people make up a significant percentage of those working in data. Ensuring our voices are heard helps reduce bias and makes data more accessible. There’s a reason so many neurodivergent people are drawn to data. It’s a field where different ways of thinking aren’t just accepted, they’re needed.</p><h3><strong>Karl’s Story: Turning Dyslexia into a Superpower in Data</strong></h3><p><strong>Karl, how has being neurodivergent influenced your work in data?</strong></p><p>My neurodivergence and success are intrinsically linked, although it took me a long time to realise that. I excel at high-level strategic thinking, simplifying complex problems, and spotting connections others might miss.</p><blockquote>Early in my career, written communication was a challenge. I used to sit there forever, paralysed with fear over sending an email. But as I progressed, I realised that my ability to write a perfect email wasn’t what made me valuable. It was my ability to think differently and innovate.</blockquote><p>I’ve now been in data for 15 years, leading enterprise digital transformation, cloud automation, and technology strategy. I joined VMO2 in August last year as a Data Culture Lead in our Data Democratisation team.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cJXYEzImwGgki1s3MZU9hA.jpeg" /><figcaption>Before joining VMO2, Karl held data leadership positions at KLM, Slack and Synaptek.</figcaption></figure><p><strong>Can you share a bit about your neurodivergence and how it has shaped you?</strong></p><p>I’m dyslexic, and for years, I hid it. It felt like an affliction, something that made me less capable than others.</p><blockquote>When you’re diagnosed, you’re often told all the things you’re not good at, and it took me nearly ten years to realise that being dyslexic, and being neurodivergent, actually makes you far superior at some things.</blockquote><p>I take longer to read things, and my writing can sometimes be misinterpreted, but I excel at seeing the bigger picture. I understand how systems and processes connect, and I recognise the nuances of how people work together.</p><p>Once I realised these were strengths, I started to lean into them, and that’s when I understood that they’re what make me good at my job.</p><p><strong>What strengths do neurodivergent individuals bring to data?</strong></p><p>We tend to be natural problem-solvers. Dyslexic people, in particular, are known for thinking in 3D, being highly creative, and challenging the status quo. We also develop strong resilience. We’ve spent our lives working around challenges, so we bring that adaptability into our careers.</p><blockquote>The world of work is changing. Employers are recognising that neurodivergent individuals bring unique strengths. I even heard on a podcast recently that GCHQ (the UK’s Intelligence &amp; Cyber Agency) actively recruits neurodivergent people for their problem-solving skills.</blockquote><p>That shift in mindset — from seeing dyslexia as a barrier to seeing it as a positive trait — has made a huge difference.</p><p><strong>What’s your advice for neurodivergent individuals considering a career in data?</strong></p><p>Get to know yourself — your strengths, and your weaknesses. Dive in and don’t be afraid, because the things that you think you’re not good at might not matter.</p><p>In data, your ability to think in 3D, to see the bigger picture, and to link systems, processes and people together is your USP.</p><p><strong>Any parting words of advice?</strong></p><p>Neurodiversity is often hidden, but it’s important we normalise these conversations and ensure our teams reflect the diversity of our customers. The more perspectives we include, the better our data insights will be.</p><p>If you think you might be neurodivergent, get an assessment. Worst case, you find out how to make <em>you </em>even better. Why wouldn’t you want to do that? Learn how to make your qualities sing.</p><p>Want to come work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace/SearchJobs/?jobRecordsPerPage=8&amp;jobOffset=8">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5045f89b593f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Levelling the Data Field: How VMO2 is bringing data literacy to the masses]]></title>
            <link>https://medium.com/@vmo2techteam/levelling-the-data-field-how-vmo2-is-bringing-data-literacy-to-the-masses-aab8c5936fda?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/aab8c5936fda</guid>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[data-literacy]]></category>
            <category><![CDATA[data-democratization]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Fri, 08 Nov 2024 17:16:30 GMT</pubDate>
            <atom:updated>2024-11-08T17:16:30.949Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qqGFlbtamjOKtyB4s6BrOQ.png" /></figure><p>At Virgin Media O2, we’re breaking down barriers to make data accessible to everyone. Our mission? To put powerful data, insights and tools into the hands of every single employee, empowering them to make smarter, data-driven decisions.</p><p>Enter Data University. It’s more than just training; it’s our bold initiative to equip every colleague with the skills and confidence to unlock their potential in a data-driven world. Here, we’re not just keeping pace; we’re leading the way.</p><p>We sat down with Tony Gibbs, our Head of Data University, to discuss what makes our approach unique, what we’re enabling, and how his own experiences — from “dying on the job” to overcoming homelessness — fuel his belief in resilience, self-improvement, and lifelong learning — all values central to Data University.</p><p><strong>What is Data University?</strong></p><p>Data University is our all-in approach to build data literacy at Virgin Media O2 by providing every single colleague with the skills they need to feel confident working with data. Our aim is to help our organisation to make smarter decisions, better understand customers, boost innovation, improve customer satisfaction and more.</p><blockquote><strong><em>It’s for everyone — from executives to front-line staff — with learning journeys tailored to the needs of all levels of the business.</em></strong></blockquote><p><strong>What is it enabling at VMO2?</strong></p><p>Data is at the core of everything we do. In today’s world of rapid change, standing still isn’t an option. Data University equips everyone at VMO2 to not only keep pace but to leap forward.</p><blockquote><strong><em>This isn’t just a nice-to-have in today’s world, where data underpins everything — it’s an essential.</em></strong></blockquote><p><strong>How does our approach differ to other companies?</strong></p><p>Many companies focus on data training for technical roles only. We’ve designed Data University to support everyone, at all levels. We’re proactive in addressing business needs, anticipating what teams need, instead of waiting for requests.</p><p>With 11 years in the field, I can honestly say this approach is different to anything I’ve encountered before.</p><blockquote><strong><em>When the opportunity came up to lead a grassroots data literacy initiative here at Virgin Media O2, I knew I wanted to be a part of it.</em></strong></blockquote><p><strong>Tell us about your role in Data University.</strong></p><p>As Head of Data University, I set the strategy to make sure our programs deliver real value across the organisation. I work closely with different teams to align our offerings with business needs and create growth opportunities for our people. From shaping learning personas to developing content and delivering training, I’m hands-on to make sure Data University equips colleagues for the future.</p><p><strong>What’s your favourite part of your job?</strong></p><p>Hands down, it’s the people. Seeing colleagues transform as they start using data to drive change and make confident decisions is incredibly rewarding. Watching them adapt to new tools and take control of their own development is hugely fulfilling. I’m excited to see where they go with these skills.</p><p><strong>What’s shaped your approach to work and life?</strong></p><p>I didn’t have a strong start. At 16, I stammered, worked a low-paying factory job, and, for a while, lived rough. During this time, I contracted Bell’s palsy, which left one side of my face paralysed. It was a dark period, but it changed my perspective.</p><p>One day, I saw an old milk bottle with layers: milk on the bottom, cream on top, and scum above that. I thought, “If scum rises, so can I.” That simple metaphor inspired me to get educated, work hard, and always push forward.</p><p><strong>Tell us one thing about yourself that not many people know.</strong></p><blockquote><strong><em>I died at work and still finished the task that killed me.</em></strong></blockquote><p>Early in my career, I was repairing electrical equipment. I’d asked a customer to unplug the machine before I started working. I heard the click and got to work, but the next thing I knew, I was waking up on the floor, confused and in pain.</p><p>I’d gotten a serious electric shock, had to be hit with a broom handle to be freed from the current, and was resuscitated because I’d turned grey. Once I came to, all I could think about was finishing the repair — despite my shaky hands.</p><p>That experience taught me how fragile life is and to be grateful for every moment.</p><p><strong>Any parting words of advice?</strong></p><p>One of my guiding principles is to resist comparison — it’s a toxic cycle. The only competition you have is with yourself.</p><p>Focus on lifelong learning, bettering yourself, and embracing the journey. Data University is here to support that journey for everyone at VMO2.</p><p>Want to work for us? <a href="https://jobs.virginmediao2.co.uk/en_US/careersmarketplace">Check out our open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aab8c5936fda" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Future of Data is Female: Abigail Beaumont and Bhagya Reddy on Leading our Data…]]></title>
            <link>https://medium.com/@vmo2techteam/the-future-of-data-is-female-abigail-beaumont-and-bhagya-reddy-on-leading-our-data-7cef7937bd8e?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/7cef7937bd8e</guid>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Tue, 08 Oct 2024 22:24:27 GMT</pubDate>
            <atom:updated>2024-10-09T23:03:28.933Z</atom:updated>
            <content:encoded><![CDATA[<h4><strong>The future of data is female: Abigail Beaumont and Bhagya Reddy on leading our Data Democratisation revolution at VMO2</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*6bRXKjqhRE2aEj_QOpV4NQ.jpeg" /><figcaption>Abigail Beaumont and Bhagya Reddy</figcaption></figure><p>At VMO2, we’re on a journey to transform our organisation into a data-driven powerhouse — providing every single colleague with access to high-quality data and tools to make informed, data-driven decisions.</p><p>At the helm in driving this seismic shift are two female visionaries and leaders, our Executive Director of Data Intelligence and Data University and Abigail Beaumont, Director of Data Strategy and Culture.</p><p>For Women in Tech Week, we caught up with them to hear about their mission to bring data to the masses, and the importance of creating an environment where women are encouraged to lead and succeed.</p><p><strong>Bhagya, why did you choose a career in data and what brought you to VMO2?</strong></p><p>Data has always fascinated me because of its power to unlock insights and solve problems at scale. It’s everywhere and impacts every decision, big or small. I started out as a data engineer, but over time I realised I wanted to be involved in shaping how data drives business strategy.</p><blockquote><strong><em>Joining VMO2 was an easy decision because of its ambitious commitment to data democratisation. Here, we’re not just making data accessible to a select few — we’re empowering everyone in the company to use it.</em></strong></blockquote><p>That’s what excited me most: the chance to contribute to a culture where data fuels innovation and decision-making at every level.</p><p><strong>Abi, how is the way we’re approaching data democratisation different to other organisations?</strong></p><p>Our approach is unique because it’s not just about access, it’s about embedding data into the company’s DNA. We’re shifting mindsets so that data becomes a part of how we think and operate.</p><blockquote><strong><em>I joined VMO2 because of the massive opportunity here to make a difference. Data is the lifeblood of any business, and VMO2’s ambitions are huge.</em></strong></blockquote><p>I wanted to be part of something that could genuinely transform the way we work and deliver value to customers.</p><p><strong>Bhagya, why is data democratisation so critical, and how are you and your team creating a data-driven culture at VMO2?</strong></p><p>The Data Democratisation programme is all about empowering employees to access and use data in their roles, regardless of their technical expertise. Traditionally, data was siloed within specific teams, but we’re changing that.</p><p>My team builds platforms that make data accessible, while Data University ensures everyone can develop the skills needed to leverage it effectively.</p><blockquote><strong><em>By breaking down barriers, we’re not just transforming how people use data — we’re creating a culture where data-driven decision-making is the norm.</em></strong></blockquote><p><strong>Abi, what advice would you give to women who want to build their careers in data?</strong></p><p>It’s important to take risks. Women often feel like they need to tick every box before applying for a role, but you don’t need to be 100% qualified to go for it. Be confident in your potential. And find mentors who can help guide you. Having people in your corner who believe in you and can offer advice is invaluable. Also, embrace opportunities to learn. Data is constantly evolving, and staying curious will keep you ahead of the curve.</p><p><strong>Bhagya, what role do you see women playing in the future of data leadership?</strong></p><p>I’m very optimistic about the future for women in data leadership. With programmes like Data Democratisation, we’re levelling the playing field and giving more women the tools and opportunities to step into leadership roles.</p><blockquote><strong><em>We need to ensure that hiring, promotion, and pay processes are equitable. It’s not just about getting women in the door — it’s about ensuring they feel supported as they move up.</em></strong></blockquote><p>Mentorship is critical. Women in leadership roles should actively support other women and help them navigate their careers.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7cef7937bd8e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cost effective telco call transcription by scaling Whisper at Virgin Media O2]]></title>
            <link>https://medium.com/@vmo2techteam/cost-effective-telco-call-transcription-by-scaling-whisper-at-virgin-media-o2-510d113c0d1b?source=rss-3a0f72e84a1d------2</link>
            <guid isPermaLink="false">https://medium.com/p/510d113c0d1b</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[telecommunication]]></category>
            <category><![CDATA[call-center]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[speech-recognition]]></category>
            <dc:creator><![CDATA[Digital at Virgin Media O2]]></dc:creator>
            <pubDate>Thu, 18 Apr 2024 12:15:52 GMT</pubDate>
            <atom:updated>2024-04-18T12:15:52.216Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*T2Vq3p0cmBcrJa6b.jpeg" /></figure><p>Virgin Media O2 is one of the UK’s largest telecommunications providers, and the scale of its contact centres is as expected, quite large.</p><p>Our call centres handle a wide range of customer requests, such as sales, account support, billing inquiries, upgrades, and technical assistance. Daily, tens of thousands of customers reach out to our Virgin Media O2 contact centres for help.</p><p>These numerous calls, amounting to millions of minutes of dialogue per month, encapsulate the essence of the customer experience. They reveal customers’ needs, their satisfaction or dissatisfaction, and contain insights into our strengths and areas for improvement. To enhance our customer service, VMO2 routinely analyses samples of call centre audio to connect with customers and identify improvement areas.</p><p>However, the approach of listening to audio recordings for insights is not scalable, particularly considering the vast amount of audio data we generate. This method offers limited perspectives on the customer experience. To gain a comprehensive understanding, it is essential to analyse thousands of call recordings.</p><p>In this article, we discuss how we developed and deployed an in-house call transcription service to transcribe our call recordings, utilising OpenAI’s open-source Whisper model, that is 20x cheaper than Google Cloud Platform (GCP) and 23x cheaper than Amazon Web Services (AWS) Speech-to-Text APIs. This work enabled us to unlock the analysis of thousands of conversations at scale using Natural Language Processing (NLP).</p><h3>1. Framing the problem</h3><p>The critical challenges we faced included:</p><p>· Acquiring the data</p><p>· Transcribing the data</p><p>· Scaling in a cost-efficient manner</p><p>Our initial step towards enabling large-scale transcription involved accessing the data. Although it may seem straightforward, the process of exporting audio files and their metadata from the third-party Contact Centre provider and importing them into GCP was one of the most time-consuming phases of this project.</p><p>Once transferred to GCP, the data consists of WAV stereo audio files paired with XML metadata files. Our task has then three processing aspects:</p><p><strong>The audio</strong>: The WAV files captures the call centre interactions. Each audio channel records a different participant; for instance, the left channel contained the call centre agent’s voice, while the right channel recorded the customer’s voice.</p><p><strong>The metadata:</strong> The XML files holds essential metadata linking the call recording to a call interaction, including timestamps, call IDs, employee IDs, system IDs, etc.</p><p><strong>The volume of data:</strong> Monthly, the call centre manages hundreds of thousands of calls, translating to tens of millions of audio minutes.</p><p>We designed and implemented a series of pipelines to ingest the audio and metadata into GCP, processed them cost-efficiently at scale, and stored the results in BigQuery (BQ) for consumption.</p><p>Now, let’s delve into the process of transcribing the audio files.</p><h3>2. The Call Transcription Speech-to-Text implementation</h3><h3>2.1 What is our goal?</h3><p>In this section, we outline our process of utilizing OpenAI’s Whisper [1] for audio transcription tailored to our specific needs. We opted for the “faster-whisper”[2] version because, it proved to be the most efficient at scale on GPUs.</p><p>Our objective was to produce a transcription that:</p><p>· Clearly distinguishes and labels the agent’s and customer’s speech.</p><p>· Includes interaction timestamps.</p><p>For instance, we aimed for a transcription that looks like this:</p><blockquote>Agent (0:00sec): “Hi! You’ve reached Virgin Media O2. This is John speaking. How can I assist you today?”</blockquote><blockquote>Customer (4:32sec): “Hello, John. I noticed on your website that the new 2GB internet service is available in my area. How do I upgrade?”</blockquote><blockquote>Agent (15:41sec): “Thank you for contacting us, John! I can process your upgrade in no time. Let’s first confirm your details.”</blockquote><h3>2.2 Transcribing stereo channels</h3><p>A key requirement was to prevent any mixing of the agent’s text with the customer’s and vice versa.</p><p>While there are several strategies for distinguishing speakers in an audio file (diarisation techniques [3,4]), we aimed to eliminate any chance of speaker transcription errors. Our strategy leverages the stereo format of the audio files, allowing us to transcribe each channel independently before merging the texts into a cohesive transcript, arranging the segments by their timestamps. This stereo channel transcription is a common practice in telephony calls and is offered by many third-party transcription vendors.</p><p>Below is an illustration of what a typical stereo audio file waveform appears like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/578/1*sp21AFLXf1zPIOONxd_bOA.png" /></figure><p>One channel contains the audio of the customer, and the other the audio of the agent. In typical call center interactions, customers may be placed on hold, resulting in periods of silence or background noise in the recordings. There periods of time can significantly increase the transcription time, and lead to model hallucinations (yes hallucinations! transformer-based speech-to-text algorithms do suffer from the infrequent hallucination).</p><p>To tackle this, we used a Voice Activity Detection (VAD) layer that identifies voice-containing segments of the waveform for transcription, significantly reducing the Whisper processing load. An example of this is shown below, where the grey areas are the parts of the waveform selected for transcription.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/692/1*bBv57hIDzuvWkACsyOkABQ.png" /></figure><p>Only the voice-active segments are used in the transcription process.</p><h3>2.3 Whisper parameters</h3><p>Tuning Whisper’s hyperparameters is essential to strike a balance between processing time, output quality, and the reduction of hallucinations. Hallucinations in speech-to-text refer to the model generating repeated words, an issue appearing mostly in noisy or low-quality audio files. The most important hyperparameters that required tunning for an optimal solution were the repetition penalty, temperature, beam size, and the ‘best of’ parameters of Whisper. Some of this needed to be carefully selected as their value can significantly slow down the inference. The inference produces ­word-level timestamps, providing us with extensive control for post-processing and reconstructing the entire conversation.</p><h3>2.4 Whisper post-processing</h3><p>In the final step of the implementation, we integrated the texts from both the left and right channels into a dialogue. This integration involves a post-processing layer designed to ensure that words merge into sentences coherently. Primarily, we leveraged timestamps and punctuation, employing standard subtitling techniques to bridge or separate word gaps and to merge or split text at punctuation points and speech gaps.</p><p>The output is a processed call transcript with a coherent flow between customers and agents, like the example below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/898/1*QVcccObsdVMUWpDSCgwI3A.png" /></figure><p>The end-to-end call transcription pipeline is shown below</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mi6cbULeuFNrhIMYaw3gzQ.png" /></figure><h3>2.5 What about quality of the output?</h3><p>The transcription’s quality depends a lot on the recording quality. We observed that transcriptions from low-quality audio files tended to have more errors, whereas high-fidelity recordings resulted in almost perfect transcriptions.</p><p>When compared with GCP and AWS Speech-to-Text APIs, our Whisper-based solution demonstrated better performance on our audio files. This was particularly evident with longer audio files, where Whisper outperformed both GCP and AWS in terms of accuracy as it reduced errors common in long audio transcriptions, such as not transcribing a substantial section of the audio. Whisper’s strategy of breaking down audio into 30-second segments, coupled with tuned hyperparameters to minimize hallucinations, significantly enhanced our ability to maintain accuracy over long durations.</p><p>The next phase to improve the transcription quality would involve fine-tuning Whisper on our dataset on manually transcribed audio files. However, this would require extensive manual transcription of hours of audio and was not undertaken. When manually assessing the quality of the outputs of hundreds of audio files we found that, despite occasional word errors, the outputs were practically useful. They provided clear insights into call reasons, intents, resolutions, and actions, fulfilling our core objective of extracting meaningful information from the calls.</p><h3>3. Productionising the solution — Pipelines and scaling up</h3><p>The productionised solution is comprised of three different pipelines.</p><p>· <strong>Data ingestion pipeline:</strong> responsible for ingesting the audio files and their corresponding metadata files from our third-party contact centre provider into a designated cloud storage bucket in GCP.</p><p>· <strong>Transcription pipeline:</strong> orchestrated by Apache Airflow, this pipeline processes the metadata files in Dataflow and stores the outputs in BQ, filters the audio files to identify the files worth transcribing, and then runs the call transcription Speech-to-Text pipeline in Dataflow. The latter component runs on GPUs and the resulting transcripts are stored in BQ.</p><p>· <strong>Data enhancement pipeline: </strong>Orchestrated in dbt, this pipeline takes the raw transcripts, the metadata and additional data sources and generates an enriched final table in BQ.<strong> </strong>This<strong> </strong>final table is made available to the wider organisation through an Analytics Hub exchange. This pipeline also incorporates monitoring and alerting as part of our Analytics Engineering team’s dbt implementation.</p><p>The end-to-end production pipeline is shown below</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UIOdzmFoSfUwn2g91hcLIg.png" /></figure><h3>3.1 The Call Transcription pipeline</h3><p>In this section, we delve deeper into the call transcription pipeline. This pipeline is designed to offload the transcription’s computational intensity to a dedicated Dataflow job running on GPUs, thereby isolating the heavy processing load from CPU-based tasks and data filtering activities.</p><p>The first stage is the metadata Dataflow pipeline, which is a job that utilizes Apache Beam for the IO and data processing of the metadata files. This job is responsible for reading, and processing and storing in BQ the metadata information, optimised by scaling up CPUs with lower RAM requirements.</p><p>The next stage involves the filtering of the audio files we will be processing. Post metadata extraction, we join on our historical call information tables to select the audio files worth transcribing. Not all files are valuable; some may contain non-essential content like hold music, or they may lack sufficient metadata to link back to specific call details. At the filtering stage, we can also implement any business logic to explicitly limit the amount of audio files we transcribe if cost becomes an issue.</p><p>Finally, the filtered audio files are being used as an input for the Speech-to-Text pipeline. This pipeline has minimal IO. The Dataflow job runs a Beam implementation of the Call Transcription Speech-to-Text implementation discussed previously, transcribes the audio, and stores the outputs in BQ. During runtime, any errors are collected in a dead-letter queue to monitor and analyse.</p><h3>3.2 Scaling up</h3><p>For the Speech-to-Text job, we consistently use 50 GPUs for its duration, finding this to be the optimal configuration for our workload. Autoscaling was not proven to be more cost-effective due to the intense GPU requirements of the task. After the initial worker spin-up and basic IO, GPU utilisation remains high throughout the process. Overall GPU utilisation for a typical job averages 94.5%, with the transcription phase of the Beam pipeline pushing this to an average of 98% (see image below).</p><p>The primary reason for this is simply how GPU-intensive is the transcription job, given the length of the audio files we process. The mean audio file duration is 12 minutes, with a substantial number of conversations being longer than 30 minutes. The IO time is considerably less than the processing time required for Speech-to-Text conversion on such lengthy files, resulting in highly efficient GPU usage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*C9tsodc0-SmVarJs__AADA.png" /></figure><h3>3.3 Beam templates</h3><p>To streamline the development and maintenance of Dataflow jobs that implement Apache Beam, such as our Speech-to-Text pipeline and NLP text classifiers, the Machine Learning Engineering team created a managed Beam template. This template handles the basic IO operations and scaling needs, providing a standardized approach across different projects. By abstracting the core functionality of the Beam codebase into a separate repository, we significantly reduce the need for boilerplate code, enhancing maintainability and consistency across our various projects.</p><h3>4. Cost Efficiency</h3><p>Hooray! We managed to do all the above! However, one might wonder why we chose to develop an in-house transcription pipeline rather than utilizing an off-the-shelf API service. The answer primarily lies in cost-effectiveness. Our in-house solution offers a substantial reduction in costs compared to commercial services.</p><p>For context, let’s compare our solution to GCP’s Speech-to-Text [5] and AWS’s equivalent services [6], which typically employ a tiered pricing model. A tiered pricing model means that, for example, the initial 250,000 audio minutes might be priced at one rate, with subsequent minutes priced differently, while GCP charges separately stereo channels. Below we’ve compiled a table comparing the costs of our in-house solution across different tiers against the pricing structures of GCP and AWS.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Cz-4JoE57csFc3RQB96vWg.png" /></figure><p>With our typical monthly workload ranging from 10 to 20 million audio minutes, our in-house solution proves to be significantly more cost-effective. It’s about 20 times less expensive than GCP and 23 times less expensive than AWS. Considering the volumes we handle, this cost differential can reduce a potential expense of $100,000 to just $5,000 — a substantial cost reduction. Even with GCP’s dynamic batching which does not guarantee when the outputs will be processed (transcription as low priority jobs), our service is still 13x cheaper than GCP for transcribing 20 million audio minutes of dual audio channel audio.</p><p>Interestingly, for smaller-scale use cases (e.g. 250,000 audio minutes), the savings are even more pronounced, with our solution being 69 times cheaper than GCP and 52 times cheaper than AWS. It’s important to note, however, that these cost comparisons can vary based on the specific distribution and length of your audio files. Our use case predominantly involves processing longer audio files, which influences the cost estimates above.</p><h3>5. Summary</h3><p>In this article, we’ve outlined the development of an in-house call transcription service using open-source Speech-to-Text libraries. We detailed the steps for setting up a transcription pipeline and discussed how to scale it efficiently with Apache Beam, Dataflow and Airflow.</p><p>Our approach allows us to transcribe millions of call audio minutes at a significantly lower cost compared to using GCP or AWS APIs, and with comparable -or even superior in some cases- performance. For our typical use case, which involves processing millions of minutes of extended call conversations, our in-house solution offers a cost advantage, being approximately 20 times less expensive than GCP and 23 times less expensive than AWS. When annualized, this cost efficiency translates into reducing potential expenses from $2 million per year to just $100,000. This substantial cost saving is a crucial factor to consider for organizations contemplating the development of such an internal tool, especially if they have the data science capabilities to support such an initiative.</p><p><strong>Authors</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/412/1*BuHLwtyG9cNIIwvPWNoVoQ.png" /></figure><p>Petros Syntelis, Principal Data Scientist</p><p>Berk Mollamustafaoglu, Principal Machine Learning Engineer</p><p>Work was performed with contributions from Alex Jones, former Principal Machine Learning Engineer at VMO2, members of the Product Engineering and Analytics Teams, especially Matt Nash in Data Democratisation.</p><h3>6. References</h3><p>1. <a href="https://openai.com/research/whisper">https://openai.com/research/whisper</a></p><p>2. <a href="https://github.com/SYSTRAN/faster-whisper">https://github.com/SYSTRAN/faster-whisper</a></p><p>3. <a href="https://lajavaness.medium.com/speaker-diarization-an-introductory-overview-c070a3bfea70">https://lajavaness.medium.com/speaker-diarization-an-introductory-overview-c070a3bfea70</a></p><p>4. <a href="https://arxiv.org/pdf/2101.09624.pdf">https://arxiv.org/pdf/2101.09624.pdf</a></p><p>5. <a href="https://cloud.google.com/speech-to-text/pricing">https://cloud.google.com/speech-to-text/pricing</a></p><p>6. <a href="https://aws.amazon.com/transcribe/pricing/">https://aws.amazon.com/transcribe/pricing/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=510d113c0d1b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>