<?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 Abhinai Srivastava on Medium]]></title>
        <description><![CDATA[Stories by Abhinai Srivastava on Medium]]></description>
        <link>https://medium.com/@abhinai?source=rss-4ea883de72c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*YXXtKt9V47UNE9HoOFqzIQ.png</url>
            <title>Stories by Abhinai Srivastava on Medium</title>
            <link>https://medium.com/@abhinai?source=rss-4ea883de72c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 13:10:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@abhinai/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[The beauty of Genetic Algorithms]]></title>
            <link>https://medium.com/@abhinai/the-beauty-of-genetic-algorithms-fde052d51677?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/fde052d51677</guid>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Mon, 07 Sep 2026 02:20:44 GMT</pubDate>
            <atom:updated>2026-09-07T02:20:44.860Z</atom:updated>
            <content:encoded><![CDATA[<p>A colleague once told me, when you have no idea how to solve a problem, you should use Genetic Algorithms. It is a class of algorithms that attempt to solve a problem by taking an inspiration from the process of natural selection.</p><p>The approach is very simple:</p><ul><li>Represent the solution to a problem as a sequence of bits or numbers. (called <strong>chromosomes</strong> or <strong>individuals)</strong></li><li>Generate an initial population of candidate solutions randomly.</li><li>Create a “fitness function” - A way to score how good an individual is.</li><li>Probabilistically pick the best solutions in the current population and combine them to create new solutions. Randomly flip some bits during this process to simulate the fact that offsprings are not an exact sum of their parents.</li><li>Repeat until we have a good solution.</li></ul><p>It’s amazing that this works. Machine learning experts will be quick to point out that the entire process looks a lot like gradient descent with a bit of randomness. True.</p><p>I am beginning to see this algorithm everywhere in daily life. The key elements are:</p><ul><li>Start with a random solution to any problem</li><li>Continuously improve upon the solution(s) you have.</li><li>Once in a while try something random. — This is very important because it prevents us from converging on a local maxima. Put other way, it allows us to discover new things we’d otherwise never try.</li></ul><p>In investing if you follow the advice of “cut your losers and double down on your winners” then it’s a genetic algorithm. Momentum investing looks a lot like genetic algorithms.</p><p>Paul Graham once said “<em>Work in two phases. Phase-1: Allow yourself to make mistakes. Phase-2: Aggressively fix mistakes</em>” (approx. statement from my recollection). I found it to be a very powerful advice because the best way to learn about something is by doing it. But if you don’t know how to do it in the first place, you’re going to have to start somewhere — do it poorly and fail — but learn how to do it better in the process and then doing it right. And that right there is very similar to the GA approach.</p><p>I am going to actively adopt this in my everyday approach to solving problems.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fde052d51677" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Came across a very old picture]]></title>
            <link>https://medium.com/@abhinai/came-across-a-very-old-picture-a93f3a25d381?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/a93f3a25d381</guid>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Sun, 12 Jul 2026 05:41:34 GMT</pubDate>
            <atom:updated>2026-07-12T05:41:34.829Z</atom:updated>
            <content:encoded><![CDATA[<p>I had just come back home after finishing college. I had to burn the roll before I could take it to get developed. Yes! This picture was taken before the era of digital cameras. Back when you had to take the film roll to a shop to get them developed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/520/1*JxLZ963RPK_k12_TGypxAw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a93f3a25d381" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Obstacle is the way]]></title>
            <link>https://medium.com/@abhinai/obstacle-is-the-way-a6775b674420?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/a6775b674420</guid>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Sat, 18 Apr 2026 01:43:01 GMT</pubDate>
            <atom:updated>2026-04-18T01:43:01.999Z</atom:updated>
            <content:encoded><![CDATA[<p>Reading this book changed my life.</p><p>I realized that setbacks and obstacles are a fact of how the world operates. They are like entropy. They are always going to happen. That is how this world works!</p><p>I’ve stopped expecting everything to go smoothly, because I’ve come to see that obstacles aren’t an interruption of the path — they are the path. For a long time, I treated every setback like a personal failure or an event of bad luck. But then, I came to see that challenges are a fundamental fact of existence, and once I started fully expecting them to appear, my entire mindset shifted.</p><p>When you stop viewing a problem as a “disaster” and start seeing it as a natural phenomenon the stress begins to evaporate. You find yourself at peace because you aren’t fighting reality anymore. Instead of being paralyzed, you simply acknowledge that this was bound to happen eventually. It removes the emotional weight and leaves you with a clear head.</p><p>The real shift happens when you move from mere acceptance to actual positivity. Since the obstacle is there anyway, we can train ourselves to find the utility in it. Every setback contains a specific opportunity — a chance to practice patience, to rethink a flawed strategy, or to develop a skill I would have otherwise ignored.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a6775b674420" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What impresses me in a job interview]]></title>
            <link>https://medium.com/@abhinai/what-impresses-me-in-a-job-interview-bde7e4356cd9?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/bde7e4356cd9</guid>
            <category><![CDATA[interview-tips]]></category>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Sat, 04 Jan 2025 21:41:16 GMT</pubDate>
            <atom:updated>2025-01-05T04:31:40.628Z</atom:updated>
            <content:encoded><![CDATA[<p>The original purpose of my blog was to journal my thoughts and learnings. But I’ve recently noticed that many people I interview for potential roles, look me up and read this blog in an effort to know how to impress me. I thought I’d make it easy for them.</p><p>Firstly, be yourself and be real. Don’t try to impress me by quoting things from my blog or my LinkedIn profile. The interview is about you, not me. I’m genuinely trying to get to know you. Your knowledge of me is irrelevant towards that purpose.</p><p>Second, if I cannot understand who you are during the allocated time, I’ll walk away confused. When I’m confused, I tend to say no. So the best way to prepare for the interview is to come up with a short and accurate description of who you are as a human being. Most of us do not understand ourselves very well so it’s important to take the time to reflect. Bottomline: Make it easy for me to get to know you.</p><p>My first question would be: “Tell me your story”. Here I’m looking for a short 5 minute version of where you grew up, what you studied, what have you worked on, what are the things you are good at and what are the things you are bad at. I’m just trying to get to know you. There is no judgement. Don’t be afraid to tell me about your failures and where you need to improve. It’ll make me like you more. We’re all human after all.</p><p>Working together is like any other human relationship. We have to be compatible, we have to like each other and we have to do great things together. These are the things I’m looking for.</p><p>I love intelligence, curiosity and passion. I love people who deeply care about their work and who are proud of what they do. People who’d go the extra mile to take care of the little things even when it doesn’t always make sense just because their pride wouldn’t let them leave something imperfect. People who love to play just because they’re curious to find out how the universe works. People who are very open about their mistakes because that’s the only way to improve. People who are willing to do just about anything and no work is beneath them because there is beauty to be found in even the most mundane tasks and because there is always a higher purpose.</p><p>These are the things I’m trying to find out during our conversation. And if you’re not this person, don’t fake it. It’s not a reflection on you. It only means we’re not compatible and that’s okay. If we get this wrong, we’d both be miserable working together and what good would that do to either of us?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bde7e4356cd9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[One shot writing]]></title>
            <link>https://medium.com/@abhinai/one-shot-writing-9a6c8ced37d4?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/9a6c8ced37d4</guid>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Sun, 10 Nov 2024 16:08:41 GMT</pubDate>
            <atom:updated>2024-11-10T16:08:41.032Z</atom:updated>
            <content:encoded><![CDATA[<p>There are two schools of thought when writing blog posts. First one dictates that you compose a blog post then keep revising it until it becomes perfect. The second one recommends writing as if you’re talking. The blog posts are imperfect but flow very naturally.</p><p>I prefer the second one. I call it one shot writing. All my blog posts are written in less than 10 minutes from start to finish and talk about exactly one topic.</p><p>I tend to chase perfection in everything in life. So why should I prefer the imperfect writing style of one-shot writing? Because from experience I’ve noticed that when I write without stopping to think, I use slightly different muscles in the brain and the end result is somehow very natural. The arguments are a bit clearer and cleaner. There is little overthinking involved.</p><p>This blog post was written in about 4 mins. :)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9a6c8ced37d4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What if I lived 200 years ago?]]></title>
            <link>https://medium.com/@abhinai/what-if-i-lived-200-years-ago-01326e6ec683?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/01326e6ec683</guid>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Sun, 11 Aug 2024 03:35:43 GMT</pubDate>
            <atom:updated>2024-08-11T03:35:43.647Z</atom:updated>
            <content:encoded><![CDATA[<p>Perhaps the hair would be longer? The uniform certainly seems very uncomfortable. I’d definitely not be smiling for this picture so that is captured well here. :)</p><figure><img alt="Abhinai Srivastava" src="https://cdn-images-1.medium.com/max/1024/1*L1tSLuN0_3S_m23iHCD6lw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=01326e6ec683" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Maturity, the frontal lobe and the magic age of 25]]></title>
            <link>https://medium.com/@abhinai/maturity-the-frontal-lobe-and-the-magic-age-of-25-2596306a1959?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/2596306a1959</guid>
            <category><![CDATA[life-thoughts]]></category>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Fri, 26 Jul 2024 08:15:05 GMT</pubDate>
            <atom:updated>2024-07-26T08:15:05.881Z</atom:updated>
            <content:encoded><![CDATA[<p>The frontal lobe is a part of the brain that is located right behind the forehead. It is responsible for speech, motor co-ordination, planning, reasoning, problem-solving, decision-making and regulating emotions.</p><p>The frontal lobe is one of the last parts of the brain to fully mature. It typically reaches full maturity in the mid-20s, around the age 25.</p><p>So 25 is the magic age that truly separates childhood from adulthood. We see examples of this all around us every day. Those of us who crossed that age remember that something changed around that time and we started feeling like adults. <strong>I remember that sometime around that age, I became “me”.</strong> I also remember that technical interview questions that felt impossibly hard before, suddenly became trivially easy to solve.</p><p>After hitting the age of 25, it somehow became easier for me to talk to anyone in the age ranges of 25–75 and have a meaningful conversation. I also suddenly started finding people below the age of 25 to be insufferable.</p><p>Most people below the age of 25 simply can’t be trusted with making significant life changing decisions. Their brains just aren’t there yet.</p><p><strong>Most countries consider 18 to be the legal age of adulthood. I think biology dictates that we should revise this number to 25.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2596306a1959" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My first modeling assignment]]></title>
            <link>https://medium.com/@abhinai/my-first-modeling-assignment-07d96027e1ad?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/07d96027e1ad</guid>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Thu, 04 Jul 2024 07:56:10 GMT</pubDate>
            <atom:updated>2024-08-11T03:26:34.352Z</atom:updated>
            <content:encoded><![CDATA[<p>This one time I was asked to be a model for an ad in a print magazine. They weren’t going to pay me but they promised to take professional pictures and share them with me. I thought that was a fair trade. Besides, I could then buy a print copy and send it to my parents to they could finally be proud of their son. Luckily, I didn’t have to buy the magazine. One day I received a package addressed to me that contained a few copies of the magazine. One of them is now proudly displayed at my parents’ house. Mission accomplished!</p><figure><img alt="Abhinai Srivastava" src="https://cdn-images-1.medium.com/max/942/1*Y8yn4T6fbdsfoMh83eKcrQ.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=07d96027e1ad" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The beauty of Cosine]]></title>
            <link>https://medium.com/@abhinai/the-beauty-of-cosine-698bc3314ebe?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/698bc3314ebe</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Mon, 22 Apr 2024 06:11:45 GMT</pubDate>
            <atom:updated>2024-05-11T07:56:05.973Z</atom:updated>
            <content:encoded><![CDATA[<p>In the early 2000s, a new kind of search engine came along. It was called Vivisimo and it returned clustered search results. Having a machine cluster results in a meaningful way was quite impressive.</p><p>Myself and a friend were determined to understand how it all worked. We worked separately to try and reproduce the beauty of Vivisimo. In the end both us got close. That’s when I first experienced the power of Cosine distance. My friend on the other hand landed on a novel approach of “Gravity” distance which re-purposed Newton’s formula for calculating the force of gravity between two bodies. Turns out, it was calculating the same thing as Cosine.</p><p>Since then I’ve used Cosine distance for a lot of things. I used it for text classification at Microsoft. The simplicity of (NN using) Cosine allowed me to focus on designing good features which then ended up making the end results better than SVM or Logistic Regression, the fashion of the day. Surprisingly, meany people made fun of me including some very smart research scientists from Microsoft Research. Except, they couldn’t beat my results.</p><p>I had no come back to people’s objections. Every research paper and study out there “proved” that SVM was the best classification approach. Cosine (NN) on the other hand was the “lowly” over simplified approach that couldn’t even be considered to be a valid classification algorithm. In the end, my algorithms were deployed because they were found to be most accurate.</p><p>Thankfully, these days with the advent of neural networks and RAG, I no longer have to spend energy convincing the world that there is something beautifully amazing about Cosine.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=698bc3314ebe" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My other blog]]></title>
            <link>https://medium.com/@abhinai/my-other-blog-11a834e77d59?source=rss-4ea883de72c------2</link>
            <guid isPermaLink="false">https://medium.com/p/11a834e77d59</guid>
            <dc:creator><![CDATA[Abhinai Srivastava]]></dc:creator>
            <pubDate>Sun, 17 Mar 2024 04:17:24 GMT</pubDate>
            <atom:updated>2024-03-17T04:17:24.049Z</atom:updated>
            <content:encoded><![CDATA[<p>My other blog is located at: <a href="https://abhinaisrivastava.blogspot.com/">https://abhinaisrivastava.blogspot.com/</a></p><p>I don’t know if it’s a good idea to move all the content in one place or leave it as is. Perhaps it doesn’t matter?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=11a834e77d59" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>