My last article of 2025 is a walkthrough of TabPFN from Prior Labs, a foundation model built for tabular data. A few months back, I read the papers, experimented with the model, and kept some notes. In this article, I have converted those notes into a high level overview of

My first article from the Federated Learning series is now live. This is the first part of a series where I explore federated learning in more depth. I start by explaining how federated learning works at a high level and walk through a simple NumPy example to build intuition.

NEW: India🇮🇳 is avoiding the fossil fuel detour 🪝 Where China built first on coal and gas, India is taking a shortcut >>> read in 5 graphics🧵

Wrapped up my second article in the federated learning series. This one focuses on Flower and how its app structure works in practice. It was fun and also a good way to clarify ideas by writing them down. Should be out soon.

The article is now live. This is part 2 of my federated learning series, where I implement federated training end to end using Flower. I borrow and adapt an example from the excellent Flower Labs course on DeepLearning.AI to clearly show what goes wrong with skewed local

The Codex app feels really smooth. I took an existing web app that I created for a hackathon and converted it into a better, more usable Android and iOS apps seamlessly. Codex not only helped migrate it but actually made it better and the end users (kids) are loving it. Its an

taste is a new core skill

The State of Machine Learning Competitions 2025 by ML Contests is out, and as always it gives a great snapshot of how ML competitions are changing and what actually worked in winning solutions last year. Some of the interesting takeaways are: • More than 390 machine learning

How Federated Learning changes model outcomes becomes very clear when you compare it to training models in isolation on split data. Here are two scenarios that I present in my article: 1️⃣ In the first, each team trains its own model on its own slice of data. 📉 The training

Going through the 2026 edition of The Missing Semester of Your CS Education and just like the previous editions, it is very thoughtfully prepared. Also, the full 2026 edition is available on YouTube and is free to watch.

The nice thing about flashcards generated in NotebookLM is that they can be exported to CSV. This means you can save them as a question bank and keep adding to it over time.

🚨 Historic milestone in India. 🇮🇳 While the world is on turbulence, India has opened it's 1st semiconductor fab manufacturing plant in Gujarat. Micron presented its first shipment of made-in-India memory modules to Dell for its laptops made in India for India. 💥

March is the month of exams in India. As I have been helping my 5th grader prepare for them I have been relying heavily on NotebookLM both for revision and for checking understanding. I am sharing some of the ways I have been using it in this article.

🚀 Our keynote speaker Parul Pandey kicks off the day with “Data Science in the Age of LLMs". The main stage is packed, the excitement is sky-high and the Python community is fully locked in as the conference begins with an electrifying start.

Ammaar Reshi Google AI Studio I built a learning app for my son using AI Studio, to help him with his hindi lessons. So I built something that felt like a personal Hindi tutor. One can uploads a chapter, gets line by line explanations, can listen and read aloud to get feedback on mistakes. There is also a

I have been seeing a lot of papers getting turned into minimal notebooks by the marimo team, so I tried it myself and it turned out pretty good. Turns out there is a whole collection of skills they’ve put together to help coding agents work with marimo, and one of them lets

Day 2 of using skrub : Cleaning dataframes with good defaults. Cleaning the dataframe is usually the first step in any data science pipeline but is also quite repetitive and cumbersome. The Cleaner from skrub handles messy data pretty well and gives you strong defaults to

Excited to partner with the marimo team for a live O'Reilly workshop on using marimo for AI and ML Development. Also, It’s completely free. 🌟 marimo is a reactive Python notebook and that completely transformed the way I work and think about notebooks. If you’re thinking

Learn from Parul Pandey how to bring agents to bear on data work using marimo pair 👇

This is such a useful idea💡.