<?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 Megha Goel on Medium]]></title>
        <description><![CDATA[Stories by Megha Goel on Medium]]></description>
        <link>https://medium.com/@goelmegha.171992?source=rss-9fd6ffc96091------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*aZr0OwsWZ0iMPZgcRSTDew.png</url>
            <title>Stories by Megha Goel on Medium</title>
            <link>https://medium.com/@goelmegha.171992?source=rss-9fd6ffc96091------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 22:41:27 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@goelmegha.171992/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[Learnings @ Virtual Grace Hopper Celebration 2020 (U.S.)]]></title>
            <link>https://medium.com/@goelmegha.171992/learnings-virtual-grace-hopper-celebration-2020-u-s-352f2aa6a9d4?source=rss-9fd6ffc96091------2</link>
            <guid isPermaLink="false">https://medium.com/p/352f2aa6a9d4</guid>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[grace-hopper-conference]]></category>
            <category><![CDATA[women-in-tech]]></category>
            <dc:creator><![CDATA[Megha Goel]]></dc:creator>
            <pubDate>Thu, 08 Oct 2020 09:21:33 GMT</pubDate>
            <atom:updated>2020-10-08T09:21:33.751Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y7xkoxN9KXGeCmhac5cRPQ.png" /></figure><p>I was fortunate enough to attend the Virtual Grace Hopper Celebration(U.S.) organized by AnitaB.org from 29th September till 3rd October 2020. GHC has brought together women leaders from industrial, academic and government communities which resulted in collaborative proposals, networking, mentoring, and increased visibility for the contributions of women in today’s technology fields, including computer science, information technology, research and engineering.</p><p>In the conference, Leading women researchers and techies presented their current work in tracks like ‘Artificial intelligence’, ‘Product Management’, ‘Entrepreneurship’, ‘Software Engineering’, ‘Diversity &amp; Inclusion’ ‘Data Science’ etc. I resonate most with the ‘Product Management’, ‘Artificial Intelligence’ and ‘Entrepreneurship’ and hence would like to share in this blog some of my takeaways and notes from some of the sessions held in the conference on same.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/520/1*-2vjc5aLh7SDJmlmOhyZIw.png" /><figcaption>Key words from various sessions in Product Management track</figcaption></figure><p><strong>Your New Product: What to Build and When to Build It : </strong>Ciara McDaniel, Senior Product Manager, Workday</p><p>I noted from the session that for creating roadmap of a new B2B product one can follow the below structure:-</p><p>1. Developing the product in <strong>3 </strong>phases where in each phase product features can be classified into 3 categories that are <strong>sales features</strong>(Key differentiators or unique selling proposition), <strong>basic features</strong>(features that are necessary so customers could use the product e.g. Login feature) and <strong>internal features</strong>(such as scaling, usage analytics which will not be used by customer but necessary to increase customer base)</p><p>2. The following diagram explains what to build and what not during each product phase:-</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-E2wMnJgfuHwptGYH0RW9g.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MNIkDOXVPcbUu5SD2laQMQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KLDDeqSKEdd72dnh5yyEfA.png" /></figure><p>Congratulations you made it!</p><p><strong>“Bonjour! Wie geht es dir? Dhanyawaad” — How Multilingual Is Your NLP Model? : </strong>Shreya Khurana, Senior Data Scientist, GoDaddy</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Rs_B1yh-mMIwVkM9ymDN6A.jpeg" /></figure><p>Natural language is constantly evolving.</p><p>The focus in talk was on building multilingual model that try to understand <strong>code-switched data</strong>(alternating between languages), <strong>underrepresented languages</strong> and <strong>transliterated data</strong>(converting words written in alphabets of one language into another) very well.</p><p>Key takeaway from the session was that the pre-trained language model <a href="https://en.wikipedia.org/wiki/BERT_(language_model)"><strong>BERT</strong></a>(Bidirectional Encoder Representations and Transformers) could be fine-tuned and used to develop multilingual detection model for various applications. It has a lot of knowledge about language, words in the language and how the words are related to each other. It can capture semantic relation and context very well because of bidirectional nature. It has been trained over wikipedia articles in different languages and is pre-trained to perform tasks like next sentence prediction and masked language model(50% data masked and then is predicted by model) and can be fine-tuned to do spam classification, POS-tagging, sentiment analysis etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5kypWerAEBoqvoM8FAkDBA.jpeg" /></figure><p><strong>Lessons Learned in Founding A Meaningful, Thriving Business — So You Can Build Yours Faster: </strong>Michelle Zatlyn, Co-founder &amp; COO<br>Cloudflare</p><p>Michelle is one of the only three women who have founded public infrastructure companies worth $1B+. The talk stresses on <strong>6 lessons</strong> from Michelle’s entrepreneurship journey which are as follows:-</p><ol><li>Solve a big, meaningful problem. (What’s the size of Market solution is addressing?, Are you going to impact 100M people?)</li><li>Get great people to come work with you.(A founding team that is diverse, more diverse teams produce better results.)</li><li>Retain all of your great people through investing in managers, good internal communication, offering career progression and employees could resonate with mission.</li><li>Build good business. It’s easy to compare businesses and difficult to compare technologies.</li><li>Chase person not valuation while picking the investors for long-term alliance.</li><li>Scale yourself. The rate at which you learn is important. Hire leadership team which is better than you.</li></ol><p>Find some useful resources shared in the one of the entrepreneurship workshop at the conference below:-</p><p><a href="https://www.talentedladiesclub.com/articles/three-easy-ways-to-test-if-your-business-idea-will-make-money/">https://www.talentedladiesclub.com/articles/three-easy-ways-to-test-if-your-business-idea-will-make-money/</a></p><p>60 Business and Finance Terms You Should Absolutely Know → <a href="https://www.fundera.com/blog/business-finance-terms-and-definitions">https://www.fundera.com/blog/business-finance-terms-and-definitions</a></p><p>Brilliance and resilience of women in technology is amazing and inspiring!</p><p>That’s all folks!</p><blockquote>Together we build future</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=352f2aa6a9d4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Learnings @ RE.WORK AI Applications Virtual Summit]]></title>
            <link>https://medium.com/@goelmegha.171992/learnings-re-work-ai-applications-virtual-summit-b09ae35e309a?source=rss-9fd6ffc96091------2</link>
            <guid isPermaLink="false">https://medium.com/p/b09ae35e309a</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[customer-success]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[fraud-detection]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Megha Goel]]></dc:creator>
            <pubDate>Tue, 22 Sep 2020 14:16:57 GMT</pubDate>
            <atom:updated>2020-09-22T14:16:57.910Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1be1HuPV6REB0TEuLZKIxw.jpeg" /></figure><p>I was fortunate enough to attend the internationally renowned conference <strong>‘AI Applications Virtual Summit’ </strong>organised by <strong>RE.WORK</strong> on 17th and 18th September 2020. This summit was attended by <strong>AI pioneers</strong> from organisations such as MasterCard, IBM, HSBC, Airbus, SAP, Nike, BBC, Google, Salesforce etc. The event brought together the latest technological advancements as well as practical examples to apply <strong>AI</strong> to solve challenges in business and society.</p><p>In this blogpost, I would like to share <strong>my takeaways and learnings from two</strong> of the many sessions that were conducted at the summit by various AI pioneers from areas like <strong>Marketing, Finance, E-commerce and Insurance</strong>.</p><h3><strong>Using Deep Learning with Word Embeddings to improve Customer Satisfaction (</strong>Eric Charton, Senior AI Director, National Bank of Canada<strong>)</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r0Qww8v65rseA0hYNF_-fg.jpeg" /></figure><p>Understanding customer satisfaction in retail banking requires exploring and comprehending multiple sources of feedback, such as emails, social networks reviews, web feedback, bot interactions, as well as speech-to-text transcripts collected from call centres. Hence, it’s important to classify and label the data in order to further analyse and derive insights out of it.</p><p>Text analytics solutions in the market are key word based for <strong>topic detection</strong> which in not suitable for <strong>topic classification </strong>in <strong>banking scenarios</strong> because classes in the data are <strong>highly unbalanced and word disambiguation</strong> creates the noise in keyword based system. The previous state of art established in 2010, where linear classifiers like SVM &amp; Logistic regression and strong feature selection obtained the best performances for topic detection. Then later on, the techniques like <strong>Embeddings(BERT) and Deep &amp; Convolution Neural Networks/GRU</strong> were applied on DEFT 2013(French cooking recipe) dataset in order to explore whether the <strong>deep learning classifier with embeddings can outperform the state of art</strong>.</p><p>In the discussion, the proposed system had a <strong>linguistic sub module along with neural module in order to eliminate the need of feature selection</strong>. The proposed system (neural sub module + linguistic sub module) when applied on customer data then the <strong>model performance turned out to be significantly improved</strong>. Finally, the labelled customer feedback has been further used to improve customer satisfaction.</p><p>The following paper could be explored for the above proposed system: <a href="https://link.springer.com/chapter/10.1007/978-3-030-47358-7_44">https://link.springer.com/chapter/10.1007/978-3-030-47358-7_44</a></p><h3><strong>Recent Developments in Deep Learning in Finance (</strong>Harshwardhan Prasad, VP — Quant Analytics, Morgan Stanley)</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7vvcuBX-X-MIkasKQs2vFA.jpeg" /></figure><p>AI applications in finance involves use-cases like <strong>risk assessment, portfolio management, Algo-trading, fraud detection, sentiment analysis and financial text mining</strong>. Cryptocurrencies, behavioural finance, blockchain, HFT and derivatives market have great potential for research.</p><p><strong>RNN based models(in particular Long Short -Term Memory model), CNN and DMLP</strong> have been used extremely in implementations. The reason for same is that the transactional data is sequential in nature like sequence of amount of money spent or time elapsed between two transactions. Also, LSTM solves vanishing gradient problem.</p><p>The research paper called ‘<strong>Deep learning detecting fraud in credit card transactions </strong>(https://ieeexplore.ieee.org/document/8374722) could be explored for <strong>fraud detection and risk assessment</strong>. In the paper, the study on fraud detection in credit card transaction indicated that LSTM and GRU model outperformed significantly the baseline ANN. There were also some other research papers that were discussed where techniques like <strong>HMM (Hidden Markov Model) and DuSVAE (Dual Sequential Variational Auto Encoder) have been used for fraud detection</strong>.The use-cases like financial text mining and portfolio management were discussed in the talk. Other aspects like transfer learning and explainability of deep learning model were shared in the discussion.</p><blockquote><strong>Artificial intelligence is powerful and is predicted to contribute massively to the global economy.</strong></blockquote><p>Before I close the blog, I would like to share that most of the talks discussed the importance of <strong>Ethics, Governance and Diversity &amp; Inclusion</strong> that are required in <strong>Artificial Intelligence development </strong>irrespective of Industry.</p><p><strong>Ethical AI</strong> ensures fairness, accountability, security and inclusivity. <strong>Governance in AI</strong> requires reinforcement of frameworks at all levels of business. <strong>Diversity and inclusion in AI </strong>helps to eliminate bias and discrimination in the model performance which can be achieved by including diverse background and experiences.</p><p>That’s all folks !</p><p>DISCLAIMER: Material shared here is in the nature of review of work done by others.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b09ae35e309a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>