<?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 Pooja Shetty on Medium]]></title>
        <description><![CDATA[Stories by Pooja Shetty on Medium]]></description>
        <link>https://medium.com/@poojaniranjanshetty?source=rss-1507ef2d32e1------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*tlbO2IB8ds8RIctgPlif4g.jpeg</url>
            <title>Stories by Pooja Shetty on Medium</title>
            <link>https://medium.com/@poojaniranjanshetty?source=rss-1507ef2d32e1------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 08:35:47 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@poojaniranjanshetty/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[Generative AI: From Chatbots to Robot Brains]]></title>
            <link>https://medium.com/@poojaniranjanshetty/generative-ai-from-chatbots-to-robot-brains-3e825e7a708f?source=rss-1507ef2d32e1------2</link>
            <guid isPermaLink="false">https://medium.com/p/3e825e7a708f</guid>
            <dc:creator><![CDATA[Pooja Shetty]]></dc:creator>
            <pubDate>Fri, 09 Oct 2026 13:01:58 GMT</pubDate>
            <atom:updated>2026-10-09T14:53:33.326Z</atom:updated>
            <content:encoded><![CDATA[<p>In 2022, a chatbot called ChatGPT reached 100 million users in just two months. That was faster than TikTok, Instagram, or any app in history. What made it so powerful? A technology called generative AI.</p><p>But generative AI isn’t just about chatting. It’s already writing code, designing drugs, driving cars, and — most importantly for the next decade — giving robots a brain.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bi6-1y0ZH1K5ikQDcMsF0g.png" /><figcaption>Coding the brain of Rotots using AI</figcaption></figure><p>Whether you’re a student, a developer, or just a curious reader, understanding generative AI is no longer optional. It’s becoming the foundation of how we work, create, and build. And the biggest shift is happening right now at the intersection of AI and robotics.</p><h3>What Is Generative AI?</h3><p>Traditional AI classifies things. It looks at an email and decides: spam or not spam. It looks at a photo and says: cat or dog. It’s a sorting machine.</p><p>Generative AI creates things. It writes essays, generates images, composes music, and writes code. It doesn’t just recognize a cat — it can draw a brand-new cat that never existed.</p><p>Think of it this way:</p><ul><li>Traditional AI is like a librarian who finds the right book for you.</li><li>Generative AI is like an author who writes a brand-new book on the spot.</li></ul><p>That’s a fundamental shift. And it’s why generative AI feels so different from everything that came before.</p><h3>How Does It Actually Work?</h3><p>Generative AI models are trained on massive datasets — billions of web pages, images, books, and conversations. They learn patterns. Then they predict what comes next.</p><p>Here’s a simple example. If you type:</p><blockquote><em>“The sky is…”</em></blockquote><p>The model predicts “blue” — because it has seen that pattern millions of times. Scale that up by a factor of billions, and you get full essays, working code, and conversations that feel human.</p><p>The magic isn’t intelligence in the human sense. It’s pattern recognition at an unimaginable scale.</p><h3>The Three Main Types of Generative AI</h3><ol><li>Text models — ChatGPT, Claude, Gemini. They write, summarize, translate, and reason.</li><li>Image models — DALL·E, Midjourney, Stable Diffusion. They turn words into pictures.</li><li>Code models — GitHub Copilot, Cursor. They write and debug software alongside you.</li></ol><p>Each of these has real-world applications. But the most exciting frontier is where they all converge: robotics.</p><p>So the hierarchy looks like this:</p><p>AI → ML → DL</p><p>AI → Machines that think → Chess engine(example)</p><p>ML → Machines that learn → Netflix recommendations(example)</p><p>DL → Machines that learn → deeplyFace recognition, ChatGPT</p><h3>How LLMs Actually Work (The Simple Version)</h3><p>You’ve probably heard the term “LLM” thrown around. It stands for Large Language Model. ChatGPT is an LLM. Claude is an LLM. They’re the engine behind the generative AI boom.</p><p>Here’s how they work, without the math.</p><p>Imagine you’re reading a sentence one word at a time. After each word, you try to guess the next one. That’s essentially what an LLM does. It’s a next-word prediction machine trained on a huge portion of the internet.</p><p>But here’s the surprising part. By getting really, really good at predicting the next word, LLMs developed the ability to:</p><ul><li>Answer questions</li><li>Write code</li><li>Summarize documents</li><li>Translate languages</li><li>Reason through problems</li></ul><p>Nobody explicitly programmed these abilities. They emerged from scale. That’s what makes LLMs so fascinating — and so unpredictable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fj9Ev-rh1OESoTDXhhCSTQ.png" /><figcaption>AI or using LLM in Robotics</figcaption></figure><h3>Why This Matters for Robotics</h3><p>Here’s the connection. If an LLM can understand language and reason about it, what happens when you connect it to a robot?</p><p>Suddenly, you don’t need to code every single movement. You can just talk to the robot.</p><p>That’s exactly what’s happening right now.</p><h3>How LLMs Are Becoming Robot Brains</h3><p>For decades, robots were dumb machines. They followed strict instructions: “Move arm 30 degrees. Pick up object at coordinates X, Y.” They couldn’t handle surprises. If a cup was slightly tilted, they failed.</p><p>Generative AI is changing that — fast.</p><p>In 2024, a robot called Figure 01 made coffee after a human asked it to. No scripted commands. Just natural language. And Google’s RT-2 model lets you say:</p><blockquote><em>“Pick up the extinct animal.”</em></blockquote><p>And the robot picks up a toy dinosaur — because it understands the concept, not just coordinates.</p><h3>The Three Gifts AI Gives Robots</h3><ol><li>Language understanding — You can say “clean the table” instead of coding every motion. The robot figures out the steps.</li><li>Vision — AI models identify objects, even unfamiliar ones. A robot can see a cup it’s never seen before and still pick it up.</li><li>Reasoning — Robots can plan steps on their own. If the first attempt fails, they adjust.</li></ol><p>This is the biggest shift in robotics in 40 years. Factories can retrain robots with words, not code. Home robots become possible. Robots can work in unpredictable environments — hospitals, disaster zones, space.</p><h3>Real-World Example: The Coffee Robot</h3><p>Figure 01 didn’t have a “make coffee” program. It had a language model that understood the request, a vision system that located the coffee maker and cup, and a planning system that sequenced the steps. All working together.</p><p>That’s generative AI in robotics. It’s not science fiction. It’s happening now.</p><p>The Bigger Picture: Where This Is Heading</p><p>Generative AI started as a text tool. Then images. Then code. Now it’s moving into the physical world.</p><p>The next five years will see:</p><ul><li>Warehouse robots that respond to spoken commands</li><li>Surgical assistants that understand verbal instructions</li><li>Home robots that learn your routines and adapt</li><li>Disaster response robots that navigate chaos without pre-programmed paths</li></ul><p>The common thread? Language. LLMs give robots a way to understand us, and in return, we can finally understand them.</p><h3>Key Takeaways</h3><ul><li>Generative AI creates new content; traditional AI classifies.</li><li>AI, ML, and DL are nested concepts — AI is the broadest, DL is the most specific.</li><li>LLMs work by predicting the next word — and that simple ability enables reasoning, writing, and coding.</li><li>Robots are finally getting smart because LLMs give them language, vision, and reasoning.</li><li>The future of robotics is conversational — you’ll talk to robots, not program them.</li></ul><h3>What’s Next?</h3><p>This is just the foundation. In the coming days, I’ll go deeper into specific tools, real robotics projects, and practical ways to get started with generative AI.</p><p>If you want a daily dose of AI explained simply — with a focus on robotics and real-world applications — follow me here on Medium.</p><p>And if this article helped you understand something new, leave a comment with the one thing that clicked for you. I read and reply to every single one.</p><p>Let’s build the future together.</p><p>#ArtificialIntelligence, # GenerativeAI, #Robotics, #Machine Learning, # BeginnersGuide</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3e825e7a708f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My day as Software Engineer in D-Nitro Team @ The KPN in Netherlands]]></title>
            <link>https://medium.com/@poojaniranjanshetty/my-day-as-software-engineer-trainee-in-d-nitro-team-kpn-5cf0cb569b75?source=rss-1507ef2d32e1------2</link>
            <guid isPermaLink="false">https://medium.com/p/5cf0cb569b75</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[kpn]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[d-nitro]]></category>
            <dc:creator><![CDATA[Pooja Shetty]]></dc:creator>
            <pubDate>Fri, 30 Sep 2022 11:54:21 GMT</pubDate>
            <atom:updated>2025-10-18T05:59:15.467Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Welcome to my humble blog! Here I would like to share my experience and day-to-day activities as a member of the D-Nitro team.</em></p><p><strong><em>Intro on my team:</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZJX9VJ2Loot-Juezy_3eyg.png" /><figcaption><strong>We accelerate Development teams to increase value delivery</strong></figcaption></figure><p>The team facilitates value enablement by providing developer productivity services to the KPN organization. This accelerates the value of the delivery of the team&#39;s applications. Our goal is to accelerate the software development lifecycle for application teams by providing a paved road of practices, ways of working, and tools. We are a team of 7 individuals with a product owner, a scrum master, three Senior software engineers, and two Software engineers. I have been a part of this wonderful team as a Software Engineer since April 2022.</p><p>These are our main tech stacks: Django, Vue JS, Jenkins, Kubernetes, Artifactory, Docker, etc. We use BitBucket to store our source code or Projects. We use JIRA to keep track of the progress of our tickets. Slack is the main medium of communication.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b8hxs4ygwZFUA-0TdV-A_g.png" /><figcaption><strong>My Team outing in Utrecht (September 2022)</strong></figcaption></figure><p>We work from the office on Mondays and Thursdays and the rest of the days we work from home. I usually go to the office around 9:45. Some people arrive much earlier, but then they leave earlier in the day. Our hours are flexible, as long as we get the work done. We usually greet each other with ‘Good Morning’ before 9:00 on Slack and we have a daily ‘Standup’ meeting for 30 mins at 10:00.</p><p>In the daily ‘Standup’ meeting we discuss about the status of the tickets using the JIRA Kanban board and we also talk about ‘operations’. ‘Operations’, or ‘ops’ for short, it is the term we use to help engineers from other divisions of KPN who have queries while using D-Nitro tools. We have a separate JIRA board for operations. Sometimes during the meeting, we also discuss on developer productivity, and more interesting ones are meetings where we discuss a product issue and come up with a solution for how to address it.</p><p>We have a 4-weekly meeting called ‘TechBase review &amp; roadmap’ from 10:00 to 12:00 on Thursday. It is the place where all the tech-base teams meet and discuss new updates in their projects, or share some valuable insights that all the teams should be aware of.</p><p>We usually go to lunch at around 12:00. We have a nice cafeteria on the ground floor of our building. Lunch-time conversations are fun and we talk a lot about various topics like food, travel, technology, etc.</p><p>We have a ‘Refinement’ meeting once or twice a month on Thursdays. During this meeting, we add/update new JIRA tickets to the ready status &amp; estimate the difficulty level of the ticket and add more information about the tickets.</p><p>We have a bi-weekly ‘Alignment’ meeting on Mondays where we discuss the order of priorities on our applications which we as a team should focus on in the coming weeks.</p><p>We have a bi-weekly ‘RETRO’ meeting on Monday afternoon where we discuss things like ‘what should we spend more/less time on’ or ‘what we must have/ nice to have/ concerns or fear’ etc. The aim of this meeting is to basically increase our team’s productivity.</p><p>I am also a part of The Pink Panther group which is a girls-only group where we meet regularly and organize events. The goal of this group is to organize Safety for women and Brave space.</p><p>Every Monday from 16:00–17:00 we have a fun team bonding event called Jennie’s hour. The event was named after one of the ex-employee of our team who initiated this idea. During this hour we usually play UNO and have a lot of fun. The day ends on a happy note and we wish each other ‘Talk to you tomorrow, have a nice evening!’.</p><p>On a typical week, I spend about 4 to 7 hours on meetings. The rest of the time I work on activities related to my tickets. These activities could be understanding the program flow, adding new features, bug fixes, testing, documenting, etc. Most of the tickets that I work on are related to the ‘backend’ and a few of them are related to the ‘frontend’.</p><p>Since joining the team, I had the chance to work on projects like ‘Console’ &amp; ‘Katka’. The ‘Console’ project provides a unified UI and API that covers a growing set of infra/tooling-related tools and applications. ‘Katka’ is an orchestration tool that allows teams in IT Companies to have a centralized way of managing and monitoring the whole development process from the moment a ticket is created up to when the outcome is available in production.</p><p>We share the Pull Requests on a private channel in slack and I get some valuable feedback. The discussions on PRs are a very good learning resource.</p><p>Overall, I am happy to be a part of a team whose members are intelligent, ambitious, and encouraging.</p><p><em>I would also encourage you to check the blogs of my colleagues: </em><a href="https://medium.com/kpn-developers/why-i-love-being-a-software-engineer-at-d-nitro-fad593960060"><strong><em>Mattias</em></strong></a><em>, </em><a href="https://medium.com/@anita.bendelja/fd15fb730fd9"><strong><em>Anita</em></strong></a><em>, </em><a href="https://www.418teapot.net/2022/09/30/life-of-a.html"><strong><em>Alex</em></strong></a><em>, </em><a href="https://medium.com/@jevonwirjono/b9d39d5cbaef"><strong><em>Jevon</em></strong></a>,<em> and </em><a href="https://medium.com/@bylalafranca/love-being-a-scrum-master-in-d-nitro-80f3dce5b37d"><strong><em>Lais</em></strong></a><em>.</em></p><p><strong><em>Thank you :)</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5cf0cb569b75" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>