<?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 Shivraj Talekar on Medium]]></title>
        <description><![CDATA[Stories by Shivraj Talekar on Medium]]></description>
        <link>https://medium.com/@shivrajtalekar101?source=rss-fd1dc9c082b4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*xjDjeRl-a50D9iGH</url>
            <title>Stories by Shivraj Talekar on Medium</title>
            <link>https://medium.com/@shivrajtalekar101?source=rss-fd1dc9c082b4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 03:14:41 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@shivrajtalekar101/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[AWS Regions, Infrastructure as Code and McAloo Tikkis]]></title>
            <link>https://medium.com/@shivrajtalekar101/aws-regions-infrastructure-as-code-and-mcaloo-tikkis-1dd84de0642d?source=rss-fd1dc9c082b4------2</link>
            <guid isPermaLink="false">https://medium.com/p/1dd84de0642d</guid>
            <dc:creator><![CDATA[Shivraj Talekar]]></dc:creator>
            <pubDate>Mon, 11 Aug 2025 11:23:15 GMT</pubDate>
            <atom:updated>2025-08-11T11:29:06.249Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bYXrxdytsxNnqJe5" /></figure><p><strong>Regions</strong>: AWS operates in 37 geographic regions around the world. There are also announced plans for several more regions in locations like New Zealand, the Kingdom of Saudi Arabia, Chile, and the AWS European Sovereign Cloud.</p><p><strong>Availability Zones (AZs)</strong>: Within these regions, there are 117 Availability Zones. Each region is designed with at least three isolated and physically separate AZs to ensure high availability and fault tolerance.</p><p><strong>Edge Locations</strong>: For content delivery and a better user experience, AWS has a global network of over 400 edge locations and 13 regional edge caches. These are part of the Amazon CloudFront content delivery network (CDN) and other services.</p><p>So, by sheer numbers, you can see that the 37 regions consist of 117 AZs. To imagine that, let’s consider McDonald’s outlets throughout the world. India has some, so does the US, and maybe other countries, excluding places like North Korea. The regions are different, in this case countries (completely isolated) from other regions.</p><h3>Why complete isolation?</h3><p>To make sure that the law of the land is complied with. Many countries have laws that prevent their data from being stored offshore. So it’s like saying: setting up a McDonald’s in the USA and in India would require a different set of paperwork.</p><p>Following a multi region and multi AZ setup, however, will ensure high availability and fault tolerance. It’s like saying that even if one of your McDonald’s outlets has a gas leak and has to be closed for an afternoon, you can still visit the other one two streets away and get your McAloo Tikki.</p><h3>Amazon Edge Locations — our street vendor cart</h3><p>Now, to elaborate on the idea of Amazon Edge Locations, we will introduce a new character to our McDonald’s business, our own street vendor cart. It will not have every item that we have on our menu back in the outlet, but it will ensure that the popular burgers and Cokes (frequently used or cached content like images and text) are provided faster with low latency. These carts can be more numerous compared to the outlets (AZs).</p><h3>Choosing a region — technical considerations</h3><p><strong>Compliance</strong> — As I wrote above, you need to follow the law of the land.<br> <strong>Feature Availability</strong> — Some features might only be rolled out in certain regions. If you switch services to a region where that particular feature isn’t available, your customer might not love it, like finding the McAloo Tikki missing in a McDonald’s in New Zealand.<br> <strong>Pricing</strong> — The costs of maintaining infrastructure vary from place to place. The changes are not always drastic, but there are variations in pricing across regions.<br> <strong>Proximity</strong> — You want your service to reach your users fast. Data travels super fast nowadays, but latency cannot be completely eliminated. You really don’t want to wait a day or two for your McAloo Tikki.</p><h3>Why McDonald’s fits this concept so well</h3><p>One more reason I chose McDonald’s specifically to explain this is how beautifully it aligns with the idea of consistent infrastructure setup.</p><p>When building a popular brand, you want most of your setup to be the same for reliability. You want your burgers (your services) to taste the same. You can trust the cook (someone handling our infrastructure setup with the AWS Management Console), but you cannot expect them to nail the ingredients (cloud configurations) perfectly every time, leaving room for errors.</p><p>So, what do you do?<br> You try to automate the process. You build a conveyor belt and make sure every process is identical, by code.</p><p>That’s where Infrastructure as Code (IaC) comes in because you can code your configurations in scripts. In simple terms, IaC means writing down all your infrastructure setup — servers, networks, databases — as code so it can be deployed, updated, or replicated automatically. It is like having the exact burger recipe and kitchen layout written down so every outlet in the world can follow it without mistakes.</p><p>This process it done with Amazon CloudFormation — With CloudFormation, you can define your infrastructure as code. You create a template that describes all the AWS resources that you want (like Amazon Elastic Compute Cloud (Amazon EC2) instances), and CloudFormation takes care of provisioning and configuring those resources for you.</p><h3>Conclusion</h3><p>This was my attempt at explaining Infrastructure as Code and setting up an AWS Region in a human friendly way.<br> If you understood even a bit of this through the McDonald’s analogy, then mission accomplished, and maybe now you’ll never look at a McAloo Tikki the same way again.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1dd84de0642d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Guide to Cloud Computing: With a Cup of Coffee]]></title>
            <link>https://medium.com/@shivrajtalekar101/the-guide-to-cloud-computing-with-a-cup-of-coffeeask-c-2290762a61b8?source=rss-fd1dc9c082b4------2</link>
            <guid isPermaLink="false">https://medium.com/p/2290762a61b8</guid>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <dc:creator><![CDATA[Shivraj Talekar]]></dc:creator>
            <pubDate>Sat, 09 Aug 2025 06:18:02 GMT</pubDate>
            <atom:updated>2025-08-09T06:48:58.747Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>The Guide to Cloud Computing: With a Cup of Coffee…</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*JGPQOvM31jAkie5g" /></figure><p><strong>Client–Server Model: The Cappuccino Edition<br></strong> Picture this. You stroll into your favorite coffee shop. You look the barista in the eye and say, “One cappuccino, please.” They nod, take your order, work their magic behind the counter, and hand over a steaming cup of happiness. You sip, you’re satisfied, and the day just feels better.</p><p>That’s essentially how the client–server model works.</p><p>You <strong>(the client)</strong> make a request. The barista <strong>(the server)</strong> processes it and sends something back. In tech terms, the client is your laptop, phone, or browser, and the server is a computer somewhere — maybe far away in a data center or even right under someone’s desk , that takes your request, does the work, and returns the result. That result could be a webpage, a file, or even a meme that makes your day.</p><p><strong>Pay Only for What You Drink<br></strong> When you buy a cappuccino, you’re not paying for the espresso machine, the rent, the barista’s salary, or the fact that they’re playing overpriced indie jazz from Spotify. You’re just paying for your drink.</p><p>Cloud computing works in the exact same way.</p><p>You don’t have to own the servers, the networking equipment, or the giant air-conditioned building where all the machines live. You simply request computing resources when you need them, and you stop paying when you don’t. If things get busy, you scale up. If it’s quiet, you scale down. This is the magic of Pay-As-You-Go.</p><p>It’s like not buying a full-blown espresso machine just to make coffee twice a month (you rent the machine time when you need it, and walk away without worrying about cleaning it afterward).</p><p><strong>How Amazon Cooked Up AWS — (Dramatization)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*acrgwQR1qXB482BI" /></figure><p>Back in the early 2000s, Amazon was still just a big online bookstore. But behind the friendly “add to cart” button, their tech team was drowning in chaos. Every new feature meant spinning up more servers, configuring databases, setting up storage, and monitoring everything from scratch. Different teams were building the same tools over and over again, which was slow, expensive, and incredibly frustrating.</p><p>And then there was the nightmare of predicting demand. Amazon had to guess how much server capacity they’d need for big events like Christmas sales. Guess too high, and millions of dollars’ worth of hardware sat unused for months. Guess too low, and the site would crash during peak shopping hours making it every engineer’s worst nightmare, and probably the reason Jeff Bezos had a permanent coffee mug in hand during December.</p><p>Somewhere between these headaches, the engineers had an idea:<br> “What if we build a solid, reusable infrastructure once and let every team use it, instead of starting from scratch each time?”</p><p>This was bigger than just servers. They wanted automated systems for storage, networking, scaling, and monitoring. A kind of “plug it in and go” platform for building any application.</p><p>And then the bigger lightbulb moment hit:<br> “If we need this, other companies probably do too… and they might pay for it.”</p><p>So Amazon turned an internal survival hack into a commercial product. In 2006, they rolled out AWS with a few core services : S3 for storage, EC2 for computing power, and SQS for messaging.</p><p>Suddenly, anyone with a laptop and an internet connection could access the same kind of computing power Amazon’s engineers had. No begging your boss for new servers. No blowing lakhs on hardware that would collect dust. No waiting weeks for IT to approve your request. Just open a browser, click a few buttons, and you’ve got a server running in minutes.</p><p>That was the start of AWS and it didn’t just make Amazon rich. It completely changed the way the world thinks about computing.</p><p><strong>Why Businesses Fell in Love with the Cloud</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6maIROzkspfiUzXu" /><figcaption>Cloud Cupid</figcaption></figure><p>AWS didn’t just give the world new tools , it changed the rules of the game.</p><ul><li><strong>From fixed to variable costs</strong>: Stop paying upfront for gear you might not use.</li><li><strong>Economies of scale</strong>: Benefit from Amazon’s massive infrastructure without owning it.</li><li><strong>No capacity guessing</strong>: Never overbuy or underbuy servers again.</li><li><strong>Speed and agility</strong>: Launch ideas faster than ever before.</li><li><strong>Cost savings</strong>: No rent, no electricity bills, no hardware maintenance.</li><li><strong>Global reach</strong>: Deploy in multiple countries without even leaving your desk.</li></ul><p><strong>AWS Global Infrastructure : The Coffee Shop Analogy<br></strong> Think of AWS like a global coffee shop chain. Instead of one massive store serving the entire planet (which would be pure chaos), they have smaller shops spread out across cities.</p><p>If the espresso machine in one shop breaks, customers just walk to the shop a few streets away. No waiting for repairs. That’s <strong>high availability (</strong>keeping your app online with minimal downtime).</p><p>Now imagine if the beans get burnt in one shop. Rather than serving bad coffee, you just route the customers to another location without missing a beat. That’s <strong>fault tolerance</strong> (your service stays alive even when something fails).</p><p>In AWS terms:</p><ul><li>A <strong>Region</strong> is like a city.</li><li>An <strong>Availability Zone (AZ)</strong> is an individual coffee shop within that city.</li></ul><p>Smart businesses never put all their coffee beans in one grinder. They spread workloads across multiple AZs, so even if one goes down, the others keep serving customers.</p><p><strong>The Takeaway<br></strong> Cloud computing isn’t some mysterious, overcomplicated tech trend. It’s just a smarter, more flexible way of doing what businesses have always done :serving customers quickly, efficiently, and without wasting resources. AWS simply took the model Amazon built for itself, made it global, and gave everyone from startups to Fortune 500 companies the same capabilities.</p><p>It’s like walking into a coffee shop and realizing you’re not just a customer , you can rent the entire setup, anywhere in the world, for as long as you need it. And when you’re done, you walk away without worrying about the cleanup.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2290762a61b8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[X to Y: Mapping My Life with Machine Learning]]></title>
            <link>https://medium.com/@shivrajtalekar101/x-to-y-mapping-my-life-with-machine-learning-06b01814cc65?source=rss-fd1dc9c082b4------2</link>
            <guid isPermaLink="false">https://medium.com/p/06b01814cc65</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[supervised-learning]]></category>
            <dc:creator><![CDATA[Shivraj Talekar]]></dc:creator>
            <pubDate>Thu, 24 Jul 2025 17:38:44 GMT</pubDate>
            <atom:updated>2025-08-09T06:22:45.414Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>X to Y: Mapping My Life with Machine Learning</strong></p><p>AI is the new electricity , like how electricity revolutionarize the dark human world which stopped working as soon as the Sun went down. AI is here to transform every field that lies under the sun.</p><p>The idea that machines can reason as good as humans was first put to test by Arthur Samuel , when he built a checkers game which could outplay its Human nemesis. Especially not just through hard-coded or predefined codes but with pattern recognition and learning on data. This was Machine Learning.</p><p>Tom Mitchell, while defining the ‘well posed problem’ stated , “A computer program is said to learn from experience E with respect to some task T &amp; some performance measure P , if its performance on T, as measured by P , improves with experience E.</p><p>In short , for example — learning = improvement in earlier predictions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*f9sFdzWPPRK15Zvl" /></figure><p>Like in the above checkers example where upon training the machine was able to defeat humans through pattern recognition was an improvement over the negative result of the previous matches from the machine perspective.</p><p>(Unrelated but important notes : Narrow tasks can be specialised by repeated &amp; heavy computation).</p><p>Then the concept of Supervised learning pops up, how we train a machine learning model with pre existing input(x) and output(y) to help it build a pattern so when we introduce a new input value (x) we can get an output based on the previously learnt pattern.</p><p>Example -</p><p>In this visualization you can see how on the horizontal X -axis we have the size in square feet corresponding to price in Dollars on the vertical Y axis (both used to train the model as it is a Supervised learning method). Here we see the straight green line which is the mapping of what you will get through when you will implement Linear Regression. Which will aid you in getting the value of the new x value , meaning if you built a new house in the same area (Mumbai) , what would the price be ? That is what we will be predicting.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*VfGpyybF1nNQ-SxB" /></figure><p>Predicting numeric values (real — continuous numbers) is what we call Regression.</p><p>Just as you might not go anywhere without your bro , regression has its bro too , its classification,</p><p>Which helps in predicting which category a particular datapoint (the new x)belongs to.</p><p>This next visualisation where we predict if a patient has cancer based on the size of tumour he/she has is a classic example of classification.</p><p>0 — Benign meaning no cancer — represented in the image by ‘X’crosses</p><p>1 — Malignant meaning cancer — represented in the image by ‘O’s</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/548/0*tyQWY9v8-gdNf4iD" /></figure><p>This is an example of binary classification as it is classified into 1 of the 2 categories . There can be numerous categories when we are working with Multi class classification , something like IPL teams categorized by the number of trophies — CSK &amp; MI falling in the 5 trophy category , 3 KKR and the rest with 1 ( i’m counting RCB) , not to forget DC , PBKS &amp; LSG who will be falling in the zero category.</p><p>As of now we were only talking about 1 x — 1 y ,</p><p>But as things grow complex we might have multiple Xs , X1 , X2 , X3 — like to predict the gender (y) of a person , we might need x1 height , x2 weight &amp; x3 age , these multiple xs we will be calling features.</p><p>(Unrelated bit interesting fact : SVM support vector machines also referred as Kernels have INFINITE FEATURES.)</p><p>So as far as we have now covered we get a basic idea that Supervised learning is to find a mapping between X &amp; Y after feeding both the X &amp; Y values.</p><p>Exciting Examples -</p><p>Automated Truck -</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KyOilkn7fUFQskYd" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*4E8dyBPwZcRdJYKd" /></figure><p>X — represented the image of the road in front of the Truck. ( Left )</p><p>Y — hand position on the steering wheel when the situation of road was x (Right)</p><p>X &amp; Y s are used to train and when there are no drivers , still we can predict the steering wheel would be and motor and mechanical functions can be used to mimic the prediction. This is a super isolated simplification of how current self driving cars work</p><p>Learning theory to be implemented when learning &amp; implementing Machine Learning:</p><p>Do we have to get more GPUs , more data or which algorithm to use. Instead of trial and errors and thinking it’s a guessing game like how you build a dream11 team when you make a team of Bangladesh vs Zimbabwe when you don’t know a single player from the team. You should learn things to core and Systemic strategy should be developed instead of believing its Black Magic and un understandable.</p><p>The instructor Andrew Ng , tells an example when he visited a company which was making something in ML for the previous 6 months and then he made him realise how it wasn’t feasible which made them realise their mistake and waste of money and resources for 6 months.</p><p>Now when we move to Unsupervised Learning , the cocktail party is a really good example , where you have 1 room , N number of People and N number Microphones in the room . and everyone trying to speak would result in overlapping voices , so we have to separate the voices after collection.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/456/0*KrAkSfyRkOzIRrvU.png" /></figure><p>Unsupervised Learning is basically without any prior knowledge of any outcomes , on the basis of inputs trying to figure out a pattern. Like before RCB won their trophy in 2025 if we would have mapped a categorization , we would have realised teams wearing REd jerseys have zero trophies , which is how clustering and pattern recognition works. Making sense out of things is its core belief.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*OOpomOo28mchkbV5.png" /></figure><p>This clustering is how we do Market Segmentation — you being identified to a certain age group and being shown a certain type of real is a good example of it.</p><p>A classic example would be how we would train a dog — “ Good dog when it does something good would encourage good behaviour and vice versa.</p><p>So this was about the Lecture 1 of CS229 , I will be referring this as my revision piece along with the notes and a way of accountability of how much i could grasp and obviously learning in public trying to make sure i complete the course , or some of you might taunt me why i haven’t posted anything after blog one so .</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=06b01814cc65" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>