<?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 Addhyan Pandey on Medium]]></title>
        <description><![CDATA[Stories by Addhyan Pandey on Medium]]></description>
        <link>https://medium.com/@addhyan?source=rss-448cd3bad062------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*gaGJf0eSf4Nap5Es.jpg</url>
            <title>Stories by Addhyan Pandey on Medium</title>
            <link>https://medium.com/@addhyan?source=rss-448cd3bad062------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 04:44:17 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@addhyan/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[Why build a data science community and not just a data science team.]]></title>
            <link>https://tech.cars.com/why-build-a-data-science-community-and-not-just-a-data-science-team-a0477c227af3?source=rss-448cd3bad062------2</link>
            <guid isPermaLink="false">https://medium.com/p/a0477c227af3</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science-teams]]></category>
            <category><![CDATA[automotive]]></category>
            <category><![CDATA[data-culture]]></category>
            <dc:creator><![CDATA[Addhyan Pandey]]></dc:creator>
            <pubDate>Thu, 13 Jun 2019 23:37:46 GMT</pubDate>
            <atom:updated>2019-06-14T17:12:51.571Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nzaIvkJi0bLmiVwtvCa_7Q.png" /><figcaption>[1] Macan, Porsche’s best selling car since 2017 and my personal favorite at the moment. Source: <a href="https://www.cars.com/research/porsche-macan-2018/">Cars.com</a></figcaption></figure><p>Ask any data science manager and they won’t deny that building a team from scratch is something they would love to do at least once in their career. I got this exciting opportunity six months after joining Cars.com. I have loved every second of it. It was when I was reflecting on my work last week that I realized that we at CARS have built more than a data science team, we have built a data science community.</p><p>You might be wondering how the two are different, so I decided to blog about it. First, what does a data science community do that a team cannot?</p><blockquote><em>The community helps drive strategy and optimization and improve data culture at a company level, something that is beyond the scope of a small data science team within a company.</em></blockquote><p>Ok, but how do you define it?</p><blockquote>A data science community consists of data practitioners and enthusiasts. It aims to nourish the practice of data-driven decision making and foster the culture of continuous learning through analysis and machine learning techniques. It also incorporates helping others learn data skills, methods and data consciousness.</blockquote><p>The following content is derived from my experience working at Cars.com for the past one year and is a reflection of our data science journey.</p><h3>Understanding business et al.</h3><p>Understanding Cars.com’s business model didn’t take me much time but understanding its revenue model did. We are a classic two-sided marketplace that helps connect car buyers and sellers. But over the past 20 years, there have been many modifications and new revenue streams that have been generated. Thus, the revenue model has become very complicated but has provided opportunities for optimization.</p><p>I have spent hours discussing this with colleagues and hearing their interpretation of the company’s business, areas of growth and its optimization potential. I then conducted multiple ad hoc analyses to understand how their interpretation relates to data and it helped me immensely in listing problems that can be addressed with data science, categorizing them based on potential long- and short-term revenue impact, and listing out the bottlenecks involved.</p><p>Although the company was in a mature state when it comes to data engineering, the machine learning (ML) demand of the data was still in its infancy. I began to engage with data and software engineering peers alongside the Product Managers. This entire exercise also helped me in building and shaping the core ML team and how we establish ourselves within the organization.</p><h3>Building the core ML team.</h3><p>Building a data science team requires some prerequisites that I’ve learned from experience. Some of them are:</p><h4>1. <strong>Support from leadership is critical.</strong></h4><p>Having executive buy-in is critical to starting a team that can deliver on both short term and long term goals. Additionally, this alignment also helps in defining and re-shaping the long-term goals. Fortunately, I was able to get enough face-time with the company’s leaders. We were all on the same page when it came to our interpretation of data science, how it works, how we can leverage it and what we need to make it fully functional. This helped me convince the organization that a team alone can’t deliver much without the whole company adopting a data science mindset and we need to build a community to flourish that culture.</p><h4>2. Having a Data Science Principles and Vision</h4><p>While our vision has evolved over the years, most of the principles have remained intact when I first framed them in September 2017. The core component was to excel at building highly predictable, rigid, multi-application* and scalable Machine Learning (ML) models. The original document can be found <a href="https://github.com/carsdotcom/ds_values/blob/master/ds_values.md">here</a>, this is also the first time we are making it public so any feedback is appreciated. If you care to have 1:1 discussions about it over coffee, feel free to DM me on <a href="https://twitter.com/addhyan_pandey">twitter</a>.</p><p>Although we had an evolving vision, our theme remained consistent, which is synonymous to the “ML minded Product and Product minded ML” phrase originally coined by Grace Huang at the KDD’18. This theme has so far been well adopted by our Product partners. Machine Learning is also reflected in each Product’s roadmap and vision. Because data science has been such an integral part of Product lifecycle, we didn’t really need a data science vision, our efforts went into making sure that Product and Data Science are aligned and work in conjunction.</p><h4>3. On-boarding process</h4><p>If you’ve read so far, you might have two questions: Why not talk about the actual hiring process and why talk about principles before hiring and building a team?</p><p>When I first came to Cars.com, the company was still defining its positioning around data science ethics. This is not uncommon — most companies haven’t answered this question for themselves. But it’s important. When we started the process of hiring, I wanted to have an answer to this question for potential candidates when they asked me about ethics, vision and values. The hiring process overall was pretty generic as we had a good group of folks from across the company to participate in it coupled with a very supportive recruiting team. Some of the key additions to the process that we as a data science team were able to introduce were 1) focusing heavily on an individual’s attitude and 2) foundational knowledge in their area of expertise and not just the knowledge about few ML algorithms and programming.</p><p>Something still was lacking. An on-boarding process.</p><p>It took me at least two weeks to understand the business and its critical components when I joined. So we built a very light-weight on-boarding process that included getting access to data, servers, platforms, crash course in business fundamentals in addition to regular training. It saved us at least two weeks per hire. We improved our process with every subsequent hire and the process as a whole was introduced and managed by one of our data scientists.</p><h4>4. Data, Platform, Continuous integration and continuous delivery frameworks (CI/CD) and Communication</h4><p>Right when I joined, the company also hired a Data Science Product Manager, who in the next 1–2 years became a key partner in making ML a success at CARS. He focused on understanding the business requirements, setting expectations, managing product lifecycle and making sure that ML is core to each product.</p><p>On the other hand, I got an opportunity to focus more on the technical stuff. We started building our core data entities to support multiple ML products. We also worked on our internal ML platform and built a CI/CD pipeline that helped us test, learn and improve our products at an incredible pace. Working in a fast-paced, high-performance environment has its pros and cons. Employees didn’t have enough bandwidth, so we had to do whatever was required to get our job done. Developers stepped in as PMs, scientists stepped in as engineers and PMs stepped in as analysts.</p><blockquote>Not every company can claim that their PMs write Python codes, SQL queries and understand common data science techniques. At Cars.com, we can.</blockquote><p>As a result of stepping in when required, our (DS team’s) engineering practices improved as well and so did our productivity. One of our data scientists recently shared her thoughts on how to <a href="https://tech.cars.com/boost-productivity-for-creating-end-to-end-machine-learning-pipeline-462c43e745e0">improve productivity</a> while building ML products. In her blog post, she shares some of the challenges she has faced.</p><p>These were challenges until we built <a href="https://github.com/carsdotcom/skelebot">Skelebot</a> (aka Skelly) that has the potential to <a href="https://tech.cars.com/skelebot-3903ae47c7b3">solve some of the</a> challenges. After continuously improving it, we finally <a href="https://github.com/carsdotcom/skelebot">open sourced</a> it. As we continue to adopt open source solutions in our day-to-day work we also pledge to contribute to the community. So, watch <a href="https://tech.cars.com/">this</a> space to learn about our python library on image processing soon.</p><h3>Collaborative projects with every organization within the company (in progress).</h3><h4>1. Successful Pilot Project with Consumer Product Organization</h4><p>You need a successful pilot to boost confidence, to reflect on your work, to celebrate and to gain more buy-in from leadership. For us, <a href="https://tech.cars.com/whats-interesting-about-hotcars-b7eda5d14176">Hot Car</a> was that success, quickly followed by our pricing algorithm.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1It0pY8wrZjMyp4C-5JmJA.png" /><figcaption>[2] Hot Car Badge helped consumers make quick decision about a vehicle they would like to purchase. The model predicts days in which a vehicle will get sold. Source: Cars.com</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*o70T9dQFTfqwNbqTKEo60Q.png" /><figcaption>[3] Our Pricing Tool. We badge vehicles based on the output of our pricing algorithm. Whereas badges are determined based on the listing price of a vehicle, the algorithm that predicts price at which a vehicle will get sold doesn’t consider the price of a vehicle as an input and therefore is tougher to game.</figcaption></figure><p>The successes of Hot Car and the pricing algorithm gave us more creative space to work on complex product problems such as: (a) rebuilding pricing, (b) relevance sort aka <a href="https://growwithcars.com/best-match/">best match algorithm</a>, (c) Review Tags, (d) Homepage Recommendations, and (e) Car-Chemistry. Today, our ML products influence at least 80% of value delivered to our customers (dealerships) in terms of Product Page Visits (VDP) and Leads.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*xzQRJoMbqS0Gzdz7UhWARw.gif" /><figcaption>[4] Best Match Algorithm delivering diverse good quality result. Source Cars.com</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KmMTe2cZc-45BhJbSYUggw.png" /><figcaption>[5] Everything on the Homepage is powered by Machine Learning. Source Cars.com</figcaption></figure><h4>2. Scale</h4><p>While building these novel solutions as products, we were also engaging with other partners in the company to learn how we can scale our products. By scale, I don’t just mean the speed at which they function but also that these algorithms should work for multiple applications. An example of such a product is our internal Natural Language Processing (NLP) framework — the one that drives our Research Pages also drives our Dealer Review Pages. That said, both the pages are consumer facing. It wasn’t simple to build such a framework because of the complications of a two-sided marketplace and the differences in data that feed these algorithms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oN3i_wcosVsMJWXYZVCXJQ.png" /><figcaption>[6] NLP driven Reviews Tags. We not only focus on ‘popular’ phrases but make it contextual and diverse in terms of sentiments.</figcaption></figure><p><strong>Our next challenge</strong>: How can we make these frameworks generic enough so that they can serve two purposes: build solutions for internal teams (such as Sales, Customer Services), and be used for more technical tasks such as feature engineering?</p><p>It’s not going to be easy, but we are up for the challenge. In fact, the experience we have gained while building review tags (Fig: 6) isn’t just technical but also industrial. Multiple dealership visits and shadowing the Sales team helped us understand the industry and its challenges first-hand.</p><p>We have already started to apply these ML/NLP frameworks to solve various problems such as taxonomy building, optimizing sales pipeline and optimizing marketing funnel.</p><h3><strong>Learnings</strong></h3><h4>1. It takes time to learn about the industry</h4><p>Having intimate knowledge of the industry and customer pain points is critical to having a seat at the table for strategy discussions. We still lag in partnering with the core business strategy and the transformation team. Don’t get me wrong, the core data science team provides frequent recommendations but it’s still not accountable to drive the company’s business strategy. It’s primarily because the team is still learning the ‘field’ knowledge of our business and auto industry, in general. We continue to partner with teams across the organization, such as Sales &amp; Brand Marketing to gain more knowledge of our unique industry.</p><h4>2. Foundational mistakes while writing job descriptions</h4><p>Formerly, all our job descriptions had Masters in a STEM degree as a mandatory requirement. But we felt this wasn’t a fair requirement and narrowed opportunities for potentially qualified candidates. As an industry, we need to do more and possibly change the requirements in our job descriptions to be more inclusive.</p><h4>3. Extra hours: when work-ethic goes awry</h4><p>One might think why putting a lot of hours in is being considered a problem. In the long run, it leads to burnout and we are here to run a marathon, not a sprint. Over the years, my team and the extended team has put long hours in order to deliver, improve and maintain several ML products. It’s important to work smart and encourage a healthy work-life balance to sustain the team’s motivation.</p><h4>4. Could’ve done a lot more with ML democratization</h4><p>Democratization is easier said than done. The data science community I speak of isn’t big enough to influence every function within the company but fortunately we are scaling fast. Improving the data culture and being more impactful through ML democratization is one of my main objectives for this quarter and we couldn’t be more excited to take the bull by the horns.</p><h3>Building the data science community was a conscious effort</h3><p>As we scaled our ML products and provided incremental growth, we got to share our excitement with the entire company. This involved doing brown bag sessions, attending business meetings and sharing our work at various town halls. This helped us in teaching the company our (data science) ways of interpreting data and using it to solve business problems.</p><p>After a few sessions, we realized that there is not only a genuine interest in learning data science within the company but also a will to solve business problem or implement ideas using ML. With an increase in engagement between ML team and rest of the organization, the company is now continuously learning to solve problems through analysis and scientific techniques. Cars.com offers access to various courses on online learning platforms, which has served as a catalyst to nourish the data culture.</p><p>The cumulative effect of all these efforts, is a community that’s taking the company forward.</p><p>Want to be part of our community? Check out our <a href="https://www.cars.com/careers/">career</a> page!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a0477c227af3" width="1" height="1" alt=""><hr><p><a href="https://tech.cars.com/why-build-a-data-science-community-and-not-just-a-data-science-team-a0477c227af3">Why build a data science community and not just a data science team.</a> was originally published in <a href="https://tech.cars.com/">Cars Commerce Technology</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What’s interesting about #HotCars?]]></title>
            <link>https://tech.cars.com/whats-interesting-about-hotcars-b7eda5d14176?source=rss-448cd3bad062------2</link>
            <guid isPermaLink="false">https://medium.com/p/b7eda5d14176</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[hot-cars]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[product-development]]></category>
            <category><![CDATA[cars]]></category>
            <dc:creator><![CDATA[Addhyan Pandey]]></dc:creator>
            <pubDate>Tue, 24 Apr 2018 20:41:41 GMT</pubDate>
            <atom:updated>2018-04-25T05:23:51.355Z</atom:updated>
            <content:encoded><![CDATA[<p>By: <a href="https://medium.com/u/448cd3bad062">Addhyan Pandey</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2QKJM85IjKeN-CV4bJntqw.png" /><figcaption>2018 Wrangler, one of my favorite off road cars.</figcaption></figure><h4>About Cars.com</h4><p>Cars.com™ is a leading two-sided digital automotive marketplace that creates meaningful connections between buyers and sellers.</p><h4><strong>Background</strong></h4><p>One of the most significant trends of 2017 was a lack of consensus around implications of important events. We saw it in politics (need I say more?), music (Drake or Kendrick?) and economics (long term impact of cryptocurrency). However, the <em>process of building consensus</em> often leads to a better understanding of the system, and in our case, we learned what makes a car HOT with the launch of Hot Car badge.</p><h4><strong>Definition</strong></h4><p>How would you define a “hot car?” Would your definition be based on certain features of a car, its popularity, expected rise in demand or whether the car was trending locally? Although those definitions are relatively easy to convey, they don’t necessarily help car shoppers make a decision on buying a car. We wanted to define it in a way that captured elements of the above mentioned definitions and we realized that they all correlate well with the selling pattern of cars. This made us conclude that a car that is ‘likely to sell soon’ would be a good way of defining a Hot Car.</p><p>Whether a car is <em>hot</em> comes down to scarcity and popularity. If a consumer knows that a car they love is going to sell quickly and that the car (and models like it) are in scarce supply, they are going to feel motivated to connect with a dealer immediately. Conversely, with a “not-hot” car, consumers have the ability to shop around and take more time researching and deciding.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AinMfJlm0fiklZwbmvuVRA.jpeg" /><figcaption><em>Left [Winner Design of Hot badge]: A search result page displaying two vehicle listings, one of them being “hot.” Right: A vehicle display page displaying information that may help users make a decision.</em></figcaption></figure><h4><strong>Algorithms, ML as a service platform, ML feature catalogue et al.</strong></h4><p>Once we decided the definition of Hot Car, framing it as a data science problem was pretty easy. We treated it as a <a href="http://papersdb.cs.ualberta.ca/~papersdb/uploaded_files/571/paper_icdm06.pdf">two-stage</a> modeling process in which we predict both number of days to sell (regression problem) and the binary output of “hot” or “not-hot” (classification problem). Our key performance metrics were <a href="https://en.wikipedia.org/wiki/Root-mean-square_deviation">rmse</a> and <a href="https://en.wikipedia.org/wiki/F1_score">f measures</a> respectively.</p><p>Building the predictive model itself and fine-tuning it didn’t take us much time. Some of the key signals to our models were around vehicle quality, its features (backup camera, sun-roof), brand value (of its make), demand, supply and seasonality (in the order of importance). After multiple iterations, an acceptable model was 70% precise on the out-of-time dataset, and we thought it was good for a Minimum Viable Product (MVP). The model-performance numbers were acceptable because a. we didn’t use any activity data (to predict HotCars) at vehicle level, so as to not impact our relevance model(s) and b. we made sure that the final model isn’t favoring any particular Make (Ford, Audi etc) or any specific dealership (National or Local).</p><p>While data scientists were focused on reducing the model’s error rate, a team of machine learning engineers built a product called Torque, our internal machine learning-as-a-service platform (MLaaS), which helped us score millions of records in minutes. Torque is a computational system/engine built on top of Cars.com’s big data infrastructure that includes our data-product (internally called Optimus Prime), which provides feature catalogue for machine learning models in near real time. It also made model monitoring easy for us because model outputs were made part of an entity (depending on the level at which a model is built) as well.</p><p>Another major benefit of having model outputs in our feature catalogue was that it made estimation of the potential impact on conversion rate of not-hot listings (that our model may have caused) easy. Other parameters we drilled into deeper included the impact on search ranks and consumer behavior as a whole.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AhAxb1mlhCe96js6wRskjw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1002/1*gjop7_IeHy9xEfnYfgWFWA.png" /><figcaption>Model Monitoring. Numbers aren’t accurate and are used for illustration purpose only. Apart from percentage of Live Hot Cars, we also monitored its interaction with our search ranks, prediction quality and error rate over time among other metrices. [Image can’t be used]</figcaption></figure><h4><strong>Strategic decisions</strong></h4><p><strong>1. Delaying the launch</strong></p><p>While we could’ve easily launched the model sooner (at least a couple of weeks), we decided to do a combined launch with <em>Torque</em> and <em>Optimus Prime </em>instead.</p><blockquote>We did this because it laid perfectly with our overall ML vision at cars; i.e. doing ethical, reproducible, scalable and robust Machine Learning.</blockquote><p>At the same time, it gave us enough buffer time for regressive model testing and analysis.</p><p><strong>2. Retraining a decent model, 48 hours before going live!</strong></p><p>Decision of delaying the launch paid off. A few days before going live with the final product, our model’s performance (in terms of precision) dropped below threshold. The entire process of gathering the training dataset, cleaning it, doing feature engineering, retraining the model, setting the new model cutoffs and finally putting it to production took us less than 12 hours.</p><blockquote>A few months ago, putting ML model to production used to take us 3–4 weeks.</blockquote><p>Today, we have five of our most important <a href="https://www.cars.com/articles/carscom-leverages-algorithm-to-redefine-best-in-class--car-search-and-shopping-experience--1420697515716/">machine learning</a> models on Torque (in production), and the plan is to expand its capabilities and redesign it a bit so that it supports cascade structure across models.</p><h4><strong>Results</strong></h4><p>When it was time for machine learning, data platforms, front-end engineering, design, A/B testing and product partners to come together at the end of the test, we found there was a lot to celebrate.</p><p>The product resulted in cracking performance across our platforms with double-digit conversion rates across desktop and mobile and double digit incremental user engagement. It also improved our consumer satisfaction ratings (tough to draw causation in this case).</p><p>We initially expected market cannibalization (that users would engage with “hot cars” at the cost of “not-hot cars”), and we weren’t wrong. New users did interact with “hot cars” more on their first visit, but after the first visit their engagement with “not-hot cars” improved as well, increasing their overall engagement by more than 25 percent on mobile.</p><blockquote>Post-celebration, the Data Science team is questioning the definition of “Hot Car” again. The idea is to actively embrace, while also challenging, industry thought leaders and best practices.</blockquote><p>Does the definition have to be the same for everyone or could it be user specific? We think users should have an option to decide their own definition of #HotCars. We can make it possible through the catalogue of our machine learning products, thereby making <em>personalization</em> the centerpiece of what we do and the products we build. Welcome to Cars.com of 2018/19!</p><h4>Sidenote</h4><p>Watch out for <a href="https://medium.com/u/158051802192">Cars.com</a> tech-blog space, we will be publishing our work on personalization. Also, <a href="http://cars.com/careers">we are hiring</a>!</p><p>Read about Hot Car badge in <a href="https://www.marketingdive.com/news/carscom-revs-up-machine-learning-to-predict-how-fast-vehicles-will-sell/521562/">news</a> and our <a href="http://investor.cars.com/news/press-release-details/2018/Carscom-Predicts-How-Quickly-Vehicles-Will-Sell-With-Launch-of-AI-Powered-Hot-Car/default.aspx">press release</a>.</p><h4>References</h4><ul><li><a href="http://ieeexplore.ieee.org/document/4053032/?reload=true">http://ieeexplore.ieee.org/document/4053032/?reload=true</a></li><li><a href="https://en.wikipedia.org/wiki/F1_score">https://en.wikipedia.org/wiki/F1_score</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b7eda5d14176" width="1" height="1" alt=""><hr><p><a href="https://tech.cars.com/whats-interesting-about-hotcars-b7eda5d14176">What’s interesting about #HotCars?</a> was originally published in <a href="https://tech.cars.com/">Cars Commerce Technology</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Top 13 of 2016 in Rap & mixed genre. #TOP2016ALBUMS]]></title>
            <link>https://medium.com/@addhyan/top-13-of-2016-in-rap-mixed-genre-top2016albums-2cce86d02d99?source=rss-448cd3bad062------2</link>
            <guid isPermaLink="false">https://medium.com/p/2cce86d02d99</guid>
            <category><![CDATA[rap]]></category>
            <category><![CDATA[music]]></category>
            <category><![CDATA[2016]]></category>
            <category><![CDATA[hip-hop]]></category>
            <category><![CDATA[review]]></category>
            <dc:creator><![CDATA[Addhyan Pandey]]></dc:creator>
            <pubDate>Thu, 12 Jan 2017 04:48:34 GMT</pubDate>
            <atom:updated>2017-01-12T21:20:57.191Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*7a94nI0w0VR6LFhjgy-1Yg.jpeg" /></figure><p>Here are the few albums I couldn’t stop listening to in 2016. Grammy predictions later!</p><p><strong>13. A Moon Shaped Pool by Radiohead</strong></p><p>“Once I start playing their music, I don’t want to stop it. I can’t pinpoint what’s so addictive about their music, but it is.” Not their best product so far but one of the best music this year. <em>Daydreaming</em> is my favourite track from this album followed by <em>Identikit</em>.</p><p><strong>12. Blonde by Frank Ocean</strong></p><p>Frank has had a hand in varied kind of music, for instance: R&amp;B (<em>Lost, Orange</em>), old school rap (<a href="https://open.spotify.com/track/0fAehaneUJdlSlWXYe1KZp">Earl’s Sunday</a>), fresh music but with an experienced player (<a href="https://open.spotify.com/track/3VeW0pO2pae144rAX2oXPP">Jay Z’s <em>Oceans</em></a>) and melancholy with pop fusion in <a href="https://www.youtube.com/watch?v=IKJ_42bkiQk"><em>Thinkin Bout You</em></a>. Tracks like <em>Nike</em> and <em>Pink + White</em> are rock solid.</p><blockquote>The album is heavily backed by electronic keyboard which is Frank’s strongest suit apart from his song writing skills.</blockquote><p>This album is not as good as his previous one but will appease his fans for the time being.</p><p><strong>11. The Colour in Anything by James Blake</strong></p><p>I expected a bit better than <em>The colour in anything</em> but James Blake didn’t disappoint his fans with this album. It’s a really long album and he has put in a lot of quality work in tracks like ‘<em>Timeless’</em> and <em>‘I need a forest fire’</em> (which is a great collaboration with Bon Iver and shows their great chemistry).</p><p>Tracks like Modern Soul, Waves know shores, My willing heart and the colour in anything are highly recommended.</p><p><strong>10. Wildflower by The Avalanches</strong></p><blockquote>If there were ever a book written on the art of sampling music, there would be a chapter written about this album.</blockquote><p>I didn’t expect them to release an album this year even after a confirmatory track with Danny Brown was released. It took them 16 years to release this album, I wouldn’t say it was worth the wait because 16 years is more than an entire career of a lot of musicians. Frankie Sinatra is my favourite track from the album and I am surprised that t’s not a BIG radio hit. The album is a good listen overall but some of the tracks like Going Home are overworked and they don’t fit well in the album. Tracks like Subways, Colours and “Because I’m me” are works of art.</p><p><strong>9. The Life Of Pablo by Kanye West</strong></p><p>My first Yeezy track was “Jesus Walk” and I can’t imagine a better way to put that song, make it so entertaining yet so spiritual and effective. JW still is my top 10 Kanye West song. Since then I’ve seen Kanye experiment with music more than any other artist of his generation and with all those experiments he has been able to entertain us. Kanye, IMO has been Hip Hop’s (not rap music) biggest influencer with his wide range of artistic abilities.</p><p>That being said, Life of Pablo didn’t disappoint me either. The gospel touch in “Ultralight beam” to the Kanye being Kanye in “Famous”, from old school touch in “Real Friends” and “No More Parties in LA” to ‘earlier Kanye’ touch in “Wolves”. This isn’t one of Yeezy’s best albums, but it was a pretty good one to kickstart 2016.</p><p>I thought a few tracks like “Highlights” were overdone but overall dope beats, dope music, dope production, dope collaborations in the entire album.</p><p><strong>8. 4 Your Eyez Only by J Cole</strong></p><p>J Cole is one of the <em>realists</em> in the game and he has the ability to produce the finest music (without features). Storytelling and conscious music are J Cole’s biggest strengths and he has played to his strengths in this album. He sounds more motivated in this album than his previous work (although his previous work is better than this one). Probably because he is narrating and sort of <em>‘doing’</em> it for someone else. I wouldn’t say J Cole has delivered a classic but it’s a good album and deserves to be in top 15 this year.</p><blockquote>J Cole fans, like myself have very high expectations when it comes to J Cole. So even if this album were rated 8.5 out of 10 by consensus, I would rate it 7 out of 10.</blockquote><p>Deja Vu, Immortal and Change are my favourite tracks from the album.</p><p><strong>7. In My Mind by BJ The Chicago Kid</strong></p><p><em>BJ The Chicago Kid </em>delivered the best of <em>Soul</em> this year. Honestly, this album is one of the best debut album I’ve heard in awhile. The earlier version of <em>Church by Chance</em> is my favourite track from the album followed by <em>The New Cupid (Kdot bias)</em>. The album very much deserves the Grammys nomination it received and I would’ve been very surprised if it hadn’t been nominated. I predict that he will win at least one Grammy this year. The simplicity of the album is it’s biggest strength and I think the album even without the collaborations from other popular artists this album would still have been great.</p><p><strong>6. We Got It From Here… Thank You 4 Your Service by A Tribe Called Quest</strong></p><p>We don’t believe you ’cause we the people … The album could not have been released at a better time. It’s great both lyrically and musically. Collaborations from André 3000,Elton John, Kendrick Lamar and Busta Rhymes were great. My top tracks were Dis Generation, Moving Backwards, We the people and Kids. This albums serves as a good bridge between old and new form of rap music. This is one of the very few albums that you can listen to multiple times in a day and still discover something new every time.</p><p><strong>5. Coloring Book by Chance the rapper</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/198/0*N6Q07I4BtV01DbIE." /><figcaption><a href="https://en.wikipedia.org/wiki/Coloring_Book_(mixtape)#/media/File:Chance_3.jpg">Source — Wiki</a></figcaption></figure><p>I was hoping this album to be in top 3 but then the album #1 happened. I remember that this album changed the mood of the Chi-town. It was playing in every car and became popular on the radio. Chance as an artist commands respect from everyone in the music industry and rightly so … to gain so much popularity without signing up with any record labels is a great achievement. His manager must be the busiest man on earth. Chance is a true Kanye <a href="http://genius.com/9143679">prodigy</a> and he has proved it with this album. It’s been Chance’s year and everyone from Chicago is proud of him.</p><p><strong>4. Blank Face LP by School Boy Q</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/191/0*43_bQbz_DCQY9gVr." /><figcaption><a href="https://en.wikipedia.org/wiki/Blank_Face_LP#/media/File:Blankfacelpcover.jpg">Source — Wiki</a></figcaption></figure><p>This album is a genre in itself. It was my favourite album of the year when it was released. <em>Q being Q </em>went too dark and probably that’s why it wasn’t a big radio hit. Tracks like Dope Dealer, Groovy Tony, Tookie Knows II and Ride Out were pretty dark depicting gang-life. ‘hotnewhiphop.com’ has the best description of this album: “This album is a psychedelic maelstrom of cautionary tales, presented in the form of acid flashbacks in Q’s mind”. My favourite track is John Muir, second to best being That Part the Black Hippy Remix. The album is TDE’s best this year and proves that Black Hippy has real talent.</p><p><strong>3. Atrocity Exhibition by Danny Brown</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/202/0*kX0HyT57tGqgurAM." /><figcaption><a href="https://en.wikipedia.org/wiki/Atrocity_Exhibition_(album)#/media/File:AtrocityExhibition.jpg">Source — Wiki</a></figcaption></figure><p>Danny Brown through this album has cemented himself as the best Detroit rapper right now. <a href="http://www.villagevoice.com/music/hes-old-missing-teeth-and-has-a-funny-haircut-hes-danny-brown-raps-biggest-deal-6641356"><em>The Hybrid</em></a> always had the potential but not many expected such a classic album from him. The sad part is that the album hasn’t been nominated for Grammys primarily because of semi-decent sales. Tracks like Lost, Pneumonia, Ain’t it funny and Rolling Stone are out of this world — perfect delivery, wonderful samples, well mixed and produced.</p><blockquote>This album has a song that features Kendrick Lamar, Ab Soul, Earl Sweatshirt alongside Danny Brown and it ain’t the best song of the album- need I say more?</blockquote><p><strong>2. Malibu by Anderson .Paak</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/316/1*na-QzvIXn-6Sux0K9n4IRg.jpeg" /><figcaption><a href="https://en.wikipedia.org/wiki/Malibu_(album)#/media/File:Anderson-Park-Malibu-Cover-Billboard-650x650.jpg">Source — Wiki</a></figcaption></figure><p>Dot <a href="http://www.npr.org/2016/01/30/464562688/anderson-paak-the-dot-stands-for-detail">stands</a> for detail in Andersan .Paak’s name and his music in this album demonstrates it. .Paak as an artist is quite talented, his live shows are one of the best in R&amp;B right now. <a href="http://www.gq-magazine.co.uk/article/anderson-paak">Reportedly</a> Dr Dre has put his money on him and the last time he seriously messed with artists they ended up being the best two lyricists/rappers of our generation’s (M &amp; Kdot). I wouldn’t want to talk too much about the album, I’d rather have you check Malibu out yourself. .Paak also has a very inspiring story, do check 88 <a href="https://www.youtube.com/watch?v=pDkbxfoxvKQ">video</a> on how he came up.</p><p>My best tracks were <em>Without You</em>, <em>Come Down</em>, <em>The Waters</em> and <em>Room in here</em>.</p><p><strong>1. Telefone by Noname</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/212/0*M_voEaaFbV7i8gr0." /><figcaption><a href="https://en.wikipedia.org/wiki/Telefone_(mixtape)#/media/File:Noname_-_Telefone.jpg">Source — Wiki</a></figcaption></figure><p>Noname is the name, remember the name!</p><p>She is the girl that you probably didn’t notice when Chance was the <a href="https://www.youtube.com/watch?v=PfaKtyWlrlM">musical guest</a> on SNL.</p><blockquote>Take “Ultralight beam” -&gt; make it raw, like how gospel music should be -&gt; improve Kanye’s verse and that is Noname for you in a sentence. Her album has 10 songs and at one point of time all her songs were on my <a href="https://open.spotify.com/user/addhyan/playlist/2CbENyK2r1dANOwyF1nI0v">Daily Rap Dose</a> playlist. My favorite Noname track is Forever. Not so surprisingly this album has put her into <a href="http://www.complex.com/music/2016/09/20-best-rappers-in-their-20s-2016">Complex</a> list of best 2016 rappers.</blockquote><p><strong>Complete List</strong></p><ul><li>Didn’t consider Vince’s and Kendrick’s album in top 10 because they weren’t a full album per se.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tazS6YDiFYrgDEo4uW5GNA.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2cce86d02d99" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ending family-feud is not good enough from Samajwadi Party, needs Congress to get over the line.]]></title>
            <link>https://medium.com/@addhyan/still-not-good-enough-from-samajwadi-party-need-to-congress-over-the-line-2065266b213c?source=rss-448cd3bad062------2</link>
            <guid isPermaLink="false">https://medium.com/p/2065266b213c</guid>
            <category><![CDATA[samajwadi-party]]></category>
            <category><![CDATA[up-elections]]></category>
            <category><![CDATA[uttar-pradesh]]></category>
            <category><![CDATA[india]]></category>
            <category><![CDATA[politics]]></category>
            <dc:creator><![CDATA[Addhyan Pandey]]></dc:creator>
            <pubDate>Tue, 10 Jan 2017 04:29:43 GMT</pubDate>
            <atom:updated>2017-01-10T20:02:13.978Z</atom:updated>
            <content:encoded><![CDATA[<p>The <a href="http://www.ndtv.com/india-news/your-father-is-very-stubborn-mulayam-singh-teased-akhileshs-daughter-1646803?pfrom=home-lateststories">news</a> of Samajwadi Party (SP) getting united may seem as a happy ending to a family drama for thousands of grass root workers and party sympathizers. It has also given a ‘feel good’ factor to millions of citizens in UP but IMO that’s not ‘enough’ to win the election of 2017.</p><p>Indian politics demands ‘drama’, which I think the Yadav’s have successfully delivered but it also demands a show of strength which SP at the moment lacks and BJP is quite strong in this department (<a href="http://timesofindia.indiatimes.com/india/prime-minister-narendra-modis-parivartan-rally-in-lucknow/liveblog/56284710.cms"><em>parivartan rally</em></a>).</p><blockquote>The next few steps for SWP should be: (1) to sit down with Congress leaders; (2) negotiate the seats; (3) and get down to business of showcasing strength via rallies while the “feel good” momentum is on.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*neATrAZ-bTjFLmxzwagEnA.jpeg" /><figcaption>Image Source: Press Trust Of India</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2065266b213c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Uttar Pradesh Elections 2017. Coalition Formula.]]></title>
            <link>https://medium.com/@addhyan/uttar-pradesh-elections-2017-coalition-formula-4571d7d814c?source=rss-448cd3bad062------2</link>
            <guid isPermaLink="false">https://medium.com/p/4571d7d814c</guid>
            <category><![CDATA[politics]]></category>
            <category><![CDATA[uttar-pradesh]]></category>
            <category><![CDATA[akhilesh-yadav]]></category>
            <category><![CDATA[up-elections]]></category>
            <category><![CDATA[india]]></category>
            <dc:creator><![CDATA[Addhyan Pandey]]></dc:creator>
            <pubDate>Tue, 10 Jan 2017 03:19:36 GMT</pubDate>
            <atom:updated>2017-01-10T03:24:51.885Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/759/1*ZqUXglqdNYJVE8Nmya2Pdg.jpeg" /></figure><p>Election results are very tough to predict even with available data (e.g., opinion polls, demographic info, census). This is because it’s tough to capture sentiments of individuals through current technical capabilities.</p><p>Well, it’s tough but not impossible! With social media trends, data and actions (likes, tweets, comments, shares, dislikes etc) it’s easy to predict the political mood of urban population. In rural areas collecting data isn’t easy. One way out is designing your surveys with certain assumptions and collecting as many data points as possible.</p><p>That’s exactly what <a href="http://economictimes.indiatimes.com/news/politics-and-nation/prashant-kishor-to-report-to-pcc-chiefs-of-punjab-and-up/articleshow/51352115.cms">Prashant Kishor</a> (PK) did for Indian National Congress (INC) in early 2016 through rallies targeting farmers and low-income groups. Prashant <a href="http://www.catchnews.com/politics-news/prashant-kishor-tells-rahul-gandhi-in-secret-report-congress-can-t-win-up-1475150728.html">reported</a> back to Rahul Gandhi (RG) that they can’t win, period. The only way they could benefit from Rahul’s multiple rallies in 2015 and 2016 was to convert the efforts into <a href="http://images.catchnews.com/uploads/images/2016/09/29/table.jpg">re-establishing</a> their presence in Uttar Pradesh’s politics. Even if INC wins 50 seats in 2017, it would be a huge success for them. This can be accomplished by forming coalition with Samajwadi Party (SP) so that Congress’s votes don’t get distributed to BJP or BSP. If Congress does join hands with Samajwadi Party they’ll have to share intelligence (data and reports), give up on CM candidacy and campaign for Akhilesh Yadav (AY). The signals from Rahul Gandhi - “<a href="http://timesofindia.indiatimes.com/city/lucknow/Rahul-Akhilesh-theek-ladka-hai-SP-seeks-apology/articleshow/53458371.cms"><em>theek ladka hai</em></a>“, Sheila Dixit - “<a href="http://indianexpress.com/article/india/sheila-dikshit-says-akhilesh-yadav-better-cm-candidate-happy-to-make-step-aside-4458989/"><em>Akhilesh, a better candidate than me</em></a>“ and Akhilesh himself “<a href="http://www.livemint.com/Politics/f0lyrtpElsTmURCbBYzZPN/SP-Congress-will-get-300-seats-from-an-UP-election-alliance.html"><em>will win more than 300</em></a>“ have been that they can fight the election together.</p><p>The other story –<em>Yadav family feud</em> has given SP the much needed <em>image improvement</em> and publicity. It has been reported in a way that portrays Akhilesh as a very clean politician who is development focused and is anti <em>gunda raj. </em>Even if the party doesn’t split up the publicity has served its purpose of rebranding. It’ll help Akhilesh, the present CM get a lot of urban swing votes.</p><blockquote>“With Rahul Gandhi’s attempt to impress rural UP and Akhilesh’s huge urban UP popularity the collation formula might just work for them unless there is any last minute scam that gets uncovered or a Modi surgical strike. We can only wait and watch!”</blockquote><p>Image source: <a href="http://encredibleindia.com/up-election/uttar-pradesh-election-2017/">http://encredibleindia.com/up-election/uttar-pradesh-election-2017/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4571d7d814c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>