<?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 Benjy on Medium]]></title>
        <description><![CDATA[Stories by Benjy on Medium]]></description>
        <link>https://medium.com/@ben_olney?source=rss-2efb976b82b3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*bNyWrmAEvkk2yNYGfHZh1Q.png</url>
            <title>Stories by Benjy on Medium</title>
            <link>https://medium.com/@ben_olney?source=rss-2efb976b82b3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 23:00:18 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ben_olney/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[Structured data from LLM responses with PydanticAI Agents]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://aiadvances.org/structured-data-from-llm-responses-with-pydanticai-agents-e66eb2d5436f?source=rss-2efb976b82b3------2"><img src="https://cdn-images-1.medium.com/max/2000/1*-VBJIEYEvXeDL5pYF2QOug.png" width="2000"></a></p><p class="medium-feed-snippet">Testing the new PydanticAI framework</p><p class="medium-feed-link"><a href="https://aiadvances.org/structured-data-from-llm-responses-with-pydanticai-agents-e66eb2d5436f?source=rss-2efb976b82b3------2">Continue reading on AI Advances »</a></p></div>]]></description>
            <link>https://aiadvances.org/structured-data-from-llm-responses-with-pydanticai-agents-e66eb2d5436f?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/e66eb2d5436f</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[pydantic]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[llm-agent]]></category>
            <category><![CDATA[agents]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Sun, 08 Dec 2024 13:26:14 GMT</pubDate>
            <atom:updated>2024-12-08T14:38:04.902Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Using Computer Vision to Fight Fraud]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://aiadvances.org/using-computer-vision-to-fight-fraud-aa18ccf3f59c?source=rss-2efb976b82b3------2"><img src="https://cdn-images-1.medium.com/max/600/1*vSBYCSt_wj_1nOHN4Rl-eA.png" width="600"></a></p><p class="medium-feed-snippet">For AI Solutions engineers designing tools for businesses</p><p class="medium-feed-link"><a href="https://aiadvances.org/using-computer-vision-to-fight-fraud-aa18ccf3f59c?source=rss-2efb976b82b3------2">Continue reading on AI Advances »</a></p></div>]]></description>
            <link>https://aiadvances.org/using-computer-vision-to-fight-fraud-aa18ccf3f59c?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa18ccf3f59c</guid>
            <category><![CDATA[fraud-detection]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-solution]]></category>
            <category><![CDATA[fraud-prevention]]></category>
            <category><![CDATA[computer-vision]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Tue, 13 Aug 2024 13:19:32 GMT</pubDate>
            <atom:updated>2024-08-13T13:20:31.731Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[OpenAI Smart Speaker with Raspberry Pi: Adding Vision and Agents]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/openai-smart-speaker-with-raspberry-pi-adding-vision-and-agents-38382b6b830f?source=rss-2efb976b82b3------2"><img src="https://cdn-images-1.medium.com/max/1209/1*Ck3fxQd4s97-oH6MrQ9gQw.png" width="1209"></a></p><p class="medium-feed-snippet">Can an open source smart speaker outclass Alexa and Siri?</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/openai-smart-speaker-with-raspberry-pi-adding-vision-and-agents-38382b6b830f?source=rss-2efb976b82b3------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/openai-smart-speaker-with-raspberry-pi-adding-vision-and-agents-38382b6b830f?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/38382b6b830f</guid>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[langchain-agents]]></category>
            <category><![CDATA[raspberry-pi]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[smart-speaker]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Fri, 05 Jul 2024 13:47:57 GMT</pubDate>
            <atom:updated>2024-07-05T13:47:57.891Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Using a JSON Agent with LangChain, LangSmith and OpenAI’s GPT-4o]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/data-science/using-a-json-agent-with-langchain-langsmith-and-openais-gpt-4o-791f57cf3b45?source=rss-2efb976b82b3------2"><img src="https://cdn-images-1.medium.com/max/2600/1*QgZmZKTBSOw3YMvVvfv-rg.png" width="5376"></a></p><p class="medium-feed-snippet">Developing a chatbot to answer questions about a JSON dataset</p><p class="medium-feed-link"><a href="https://medium.com/data-science/using-a-json-agent-with-langchain-langsmith-and-openais-gpt-4o-791f57cf3b45?source=rss-2efb976b82b3------2">Continue reading on TDS Archive »</a></p></div>]]></description>
            <link>https://medium.com/data-science/using-a-json-agent-with-langchain-langsmith-and-openais-gpt-4o-791f57cf3b45?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/791f57cf3b45</guid>
            <category><![CDATA[json]]></category>
            <category><![CDATA[langchain-agents]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[langsmith]]></category>
            <category><![CDATA[gpt-4o]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Mon, 01 Jul 2024 22:40:27 GMT</pubDate>
            <atom:updated>2024-07-01T22:40:27.063Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[OpenAI Smart Speaker with Raspberry Pi]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/geekculture/openai-smart-speaker-with-raspberry-pi-5e284d21a53e?source=rss-2efb976b82b3------2"><img src="https://cdn-images-1.medium.com/max/2600/1*gMNlKaye08oO5vemLTfnpQ.png" width="4032"></a></p><p class="medium-feed-snippet">Making an -actual- Smart Speaker</p><p class="medium-feed-link"><a href="https://medium.com/geekculture/openai-smart-speaker-with-raspberry-pi-5e284d21a53e?source=rss-2efb976b82b3------2">Continue reading on Geek Culture »</a></p></div>]]></description>
            <link>https://medium.com/geekculture/openai-smart-speaker-with-raspberry-pi-5e284d21a53e?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/5e284d21a53e</guid>
            <category><![CDATA[smart-speaker]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[speech-recognition]]></category>
            <category><![CDATA[text-to-speech]]></category>
            <category><![CDATA[raspberry-pi]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Thu, 01 Feb 2024 05:34:31 GMT</pubDate>
            <atom:updated>2024-07-05T21:11:56.298Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Exploring the use of AI and Machine Learning for Mortgage Recommendations.]]></title>
            <link>https://medium.com/geekculture/exploring-the-use-of-ai-and-machine-learning-for-mortgage-recommendations-95a22ea07d13?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/95a22ea07d13</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[mortgage]]></category>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[api]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Mon, 02 Aug 2021 14:50:32 GMT</pubDate>
            <atom:updated>2021-08-03T14:39:20.735Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ChGPYf9SHo4PLTTgejvm7Q.png" /><figcaption>My latest project SwitchMortgageRates.com</figcaption></figure><h3>MortgageTech &amp; API’s:</h3><p>The year of the mortgage API was not to be in 2020. Lenders buddied up with selected partners, helping their biggest and most profitable accounts release API solutions via their own bespoke systems. The door was closed for everyone else. <strong>That is not how we brew MortgageTech innovation</strong>. Now, deep into 2021, we still await signs of a more open and transparent API integration with all major banks and building societies on the mortgage front.</p><p>The frustration waiting around for API’s made me focus on other solutions for MortgageTech. What could I work on with my self-taught knowledge of Python and Full Stack development skills in the meantime? Something that is not heavily reliant on lender integration that is. AI and machine learning was the answer.</p><h3>The Project: AI and Machine Learning for Mortgage Recommendations.</h3><p>Firstly, I wanted to design and build a website from the ground up using Python. Python is a high level programming language capable of machine learning and artificial intelligence. My main website currently relies too much on third party services. Wordpress for example, its a great blogging site but it’s simply not flexible enough to integrate more powerful code.</p><blockquote><em>So I built </em><a href="https://www.switchmortgagerates.com/"><em>SwitchMortgageRates.com</em></a></blockquote><p>The primary function right now is a lender focused API — but with a difference. The API will retry different searches if the first one fails, scanning for specific parameters within the mortgage product information. It is a Python based solution to a problem: Mortgage products are currently poorly categorised by mortgage lenders and by Twenty7Tech. T7T are currently the only API sourcing solution for mortgage products across the entire industry — which I find bizarre now in 2021!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r3rH4yRM4ViNDdMHJiZw1A.png" /><figcaption>An API result for a non-regulated buy to let switch rate mortgage option (product transfer) with BM Solutions / Birmingham Midshires. Information correct as of 02/08/2021 but should not be relied upon after this date.</figcaption></figure><p>The API is part one of the project which is now complete. <a href="https://www.switchmortgagerates.com/">SwitchMortgageRates.com</a> works well to access and retrieve product transfer mortgage deals for existing customers, that is, when the product information is provided.</p><p><strong>Part two</strong> (under development) will be testing the input from the customer to recognise and understand their goals now and in the future. It will allow customers themselves to test various mortgage and property scenarios using AI.</p><p><strong>Here’s an example:</strong></p><blockquote><strong><em>Customer A:</em></strong><em> We are looking to switch rates but might possibly move in less than 2 years time when we are planning to get married and start a family. What is the best option for us now?</em></blockquote><p><strong>AI will kick in</strong>. It will analyse their current interest rate against the proposed best new deals right now, the initial rate periods, and the customer’s plans for the future. Based on these conditions, and knowing all of the costs involved now and in the future, the recommendation will account for this scenario. Early Repayment charges will be factored in along with upfront fees and other costs. I’m even looking at involving <a href="https://www.bankofengland.co.uk/boeapps/database/FromShowColumns.asp?Travel=NIxAZxI1x&amp;FromCategoryList=Yes&amp;NewMeaningId=RSVRM&amp;CategId=6&amp;HighlightCatValueDisplay=Standard%20variable%20rate%20mortgage">standard variable rate analysis</a> from figures provided by the Bank of England.</p><p>It’s true to say that the regulator will have their own view on any potential AI capability — which is why I am testing this privately for now. So, before we get too carried away, it might be many years before this type of technology can be implemented. It may also need to be released with a big disclaimer but the concept is exciting. The machine learning part will kick in if the customer decides to save their profile. The system will remember the customer’s intentional plans and learn to adapt if these plans change, therefore also updating the AI.</p><p>For now though I’m just pleased with the result of my latest project. My first full stack Python website with some great tools for customers looking to switch mortgage rates now or in the future. The foundations are in place, ready to take on the future of AI, Machine Learning and of course MortgageTech - when the banks and building societies decide to fully open up their mortgage API’s.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=95a22ea07d13" width="1" height="1" alt=""><hr><p><a href="https://medium.com/geekculture/exploring-the-use-of-ai-and-machine-learning-for-mortgage-recommendations-95a22ea07d13">Exploring the use of AI and Machine Learning for Mortgage Recommendations.</a> was originally published in <a href="https://medium.com/geekculture">Geek Culture</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Open Banking — Greater fraud prevention and affordability analytics.]]></title>
            <link>https://medium.com/@ben_olney/open-banking-greater-fraud-prevention-and-affordability-analytics-75586ca793eb?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/75586ca793eb</guid>
            <category><![CDATA[mortgage-broker]]></category>
            <category><![CDATA[open-banking-api]]></category>
            <category><![CDATA[open-banking]]></category>
            <category><![CDATA[mortgage]]></category>
            <category><![CDATA[fintech]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Fri, 07 Feb 2020 16:01:01 GMT</pubDate>
            <atom:updated>2021-07-29T16:15:26.745Z</atom:updated>
            <content:encoded><![CDATA[<h3>Open Banking — Greater fraud prevention and affordability analytics.</h3><p>Now there’s a statement you would not normally choose to put together, Open Banking and Fraud Prevention. But now, with Open Banking we have the ability to achieve more in the fight against fraud. It’s not just about the ability to access a bank account via API to prove that it exists. There is a higher level of fraud prevention and affordability analytics thanks to complex and insightful reporting tools driven from Open Banking API’s.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EZcVtpJ5e7TRnruqrOb8_A.png" /><figcaption>Open Banking</figcaption></figure><h3>Enriched Reporting via Open Banking</h3><p>There are a number of startups now offering Open Banking API access. The level of detail in their reporting is mind blowing. These tools are aimed at providing insights into affordability, but they also offer insights that assist with fraud prevention. Good firms don’t just access bank accounts and obtain streams of raw transactional data. They process and sort the data into complex, insightful reports.</p><p>You can obtain warning indicators for Gambling occurrences and Cash Withdrawals, as well as reports on recurring sources of income over time (12 months to be precise). There are even statistics on the predicted longevity and consistency of incoming payments and charts on spend by category. It’s like Monzo for the professional advisor.</p><h3>Protection for Intermediaries</h3><p>As brokers and professional mortgage advisors we need these tools to help prevent fraud. The <a href="https://www.cifas.org.uk/insight/fraud-risk-focus-blog/mortgage-fraud-on-rise">number of fraudulent applications are on the rise</a>. One example of fraud is ‘Staged Income’. This is where an applicant will claim they have a job (typically a second job) and can evidence this income over 3 months on their bank statements. The applicant may have recently started this job in order to trick the mortgage lender into lending them more money and they are likely to stop this job once the mortgage has gone through.</p><p>With Open Banking, more of these applicants could be detected given the advanced reporting tools available. Ultimately, we always try and help our customers achieve the best mortgage they can. However, we can be struck off the panel with a mortgage lender in an instant and never be allowed to do business with them again over just one applicant trying to cheat us and the system.</p><p>Open Banking and new level reporting tools will help us demonstrate to the lenders that we are at the top of our game when it comes to helping prevent fraudulent applications and delivering comprehensive affordability assessments.</p><h3>Confidence and Faster Applications for Customers</h3><p>For customers, this technology will bring an increased confidence on the likelihood of final acceptance and inevitably speed up the time it takes to complete the initial application checks.</p><p>It could well open the door to new mortgage options for borderline applicants and First Time Buyers. This is thanks to the greater level of detail in the reporting tools that project liquidity, future outgoings and flexible expenditure.</p><p>Customers will also no longer have to try and figure out how to organise and download their bank statements online. For some customers this can be a complex and frustrating process with <a href="https://glow.mortgage/how-to-download-or-print-your-bank-statements-from-natwest/">several unnecessary steps involved</a> in their online banking dashboard.</p><p>Challenger banks like Monzo have already capitalised on making Open Banking easily accessible. Open Banking access can be approved via the <a href="https://monzo.com/blog/2018/06/14/what-is-open-banking">Monzo app</a> within a few taps, without the need to remember your online banking ID or online banking password.</p><h3>Now it’s down to the mortgage lenders:</h3><p>We are hoping that lenders will jump on board quickly with the reporting statements and summary PDF’s that Open Banking providers make available. They offer a much greater insight into the applicant than ever before. Rather than lenders applying their own standard ‘one-size-fits-all’ affordability algorithm, lenders will be able to provide more bespoke affordability assessments.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=75586ca793eb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The world of the Online Mortgage Broker]]></title>
            <link>https://medium.com/@ben_olney/the-world-of-the-online-mortgage-broker-4d4bf1c55b7?source=rss-2efb976b82b3------2</link>
            <guid isPermaLink="false">https://medium.com/p/4d4bf1c55b7</guid>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[mortgage]]></category>
            <category><![CDATA[online-mortgage-broker]]></category>
            <category><![CDATA[mortgage-broker]]></category>
            <dc:creator><![CDATA[Benjy]]></dc:creator>
            <pubDate>Thu, 07 Nov 2019 09:53:45 GMT</pubDate>
            <atom:updated>2020-01-14T12:23:58.404Z</atom:updated>
            <content:encoded><![CDATA[<h3>The World of the Online Mortgage Broker</h3><h3>Disruptor</h3><p>I could never accept that as a mortgage advisor my day would be spent booked up with face-to-face appointments, some appointments lasting more than 2 hours. Mostly organised by the estate agent, I felt bad for the folks who were forced to take time off work or even schedule a holiday day just to see me. It just didn’t make sense.</p><p>This was back in 2008. Once I gained enough experience, I ventured out from my employed mortgage advisor role with a view to make things easier. The only problem was, back then, there were very little online tools available. I ended up trialling different tools such as Skype. I installed a very fancy script on the first website I designed, once the script was clicked it would open the Skype app and start a video mortgage appointment. Customers were simply not ready for this!</p><p>That was the start of my tinkering. It was October 2017 when I officially launched <a href="https://glow.mortgage/">Glow</a>. My aim was simple, to create a truly online mortgage broker powered by experience, tech and customer service. The tech part was important, but so too was the need to match tech with great customer service.</p><figure><img alt="Glow.Mortgage Website Homepage" src="https://cdn-images-1.medium.com/max/1024/1*aDEKLnm3UimMw9IPnXXrRg.png" /></figure><p><a href="https://youtu.be/2c8uK0kfUYY">https://youtu.be/2c8uK0kfUYY</a></p><h3>Constant Evolution</h3><p>The beauty of being a disruptor means that I can shift and ship new technology in less than a day. There are a plethora of tools out there to assist with online mortgage broking if you don’t fancy developing them yourself, whether that be live chat tools or backend CRM. Automation is key, automation software allows you to process enquiries without the admin. These tools are constantly evolving and linking with other software to simplify processes, so that even a broker with basic tech skills can reduce their admin workload.</p><p>I can’t count the amount of times that I have changed, or fully u-turned a project because I have found a better solution. Constant re-assessment and evolution is key.</p><h3>Simplifying the Mortgage Quote</h3><p>In 2019 I was proud to reveal ‘Mortgage Cards’ by Glow. Mortgage cards is the most simple way to get a quote, without adverts and without clutter in less than 60 seconds.</p><figure><img alt="The start of the online quote engine." src="https://cdn-images-1.medium.com/max/1024/1*pkfyF9zdxk-jM0HWbx22ag.png" /><figcaption>Mortgage Cards by Glow</figcaption></figure><p>We looked at other quote websites and found that they either overwhelm you with choice, graphics, popups, or gimmicky speel. Then at the end of the process you have no instant connection with a mortgage broker or advisor.</p><p>With Glow, if you just want a quote, you can get one quickly. If you want to chat instantly with an experienced advisor right away, you can. The quote process is aided by live chat as well if you get stuck, or if you need help along the way you can click ‘Get Help’ in the menu.</p><figure><img alt="The last question of the online quote engine before mortgage rates are presented on screen." src="https://cdn-images-1.medium.com/max/1024/1*lO2UTW9a2F3g-z1TOKZsXA.png" /><figcaption>Compare rates clutter free in less than 60 seconds.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1N4FLQ_9YWXtGKrVFLBECA.png" /><figcaption>Mortgage Cards by Glow. Interest Rates and financial information cannot be displayed for compliance purposes.</figcaption></figure><h3>What’s Next?</h3><p>Next up we are looking to streamline the application process itself by furthering our API capabilities and we are currently looking at partners for digital conveyancing. Some lenders have over 100 questions in their mortgage application and not all questions are the same from lender to lender. This poses quite a challenge but we are already working on a solution.</p><p>We believe that 2020 will bring further open banking and integration possibilities to intermediaries. We are looking to take advantage of the services that continue to make the journey quicker, easier and less complicated for anyone looking for a mortgage.</p><p>The problem that independent firms will face in 2020 is that no sooner than a FinTech solution firm announces it’s entry into the market than it becomes aligned to a network, mortgage club or large brokerage. Thus being customised to that particular avenue of operation and losing it’s USP.</p><p>Fear not disruptors, if you have the vision then I recommend the bespoke route. Take the best of everything you see and create a truly independent solution for your own firm.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4d4bf1c55b7" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>