Blog

less than 1 minute read

2026

They'll Spend Anything

Aug 1, 2026

Except on anything ambitious. A rant about India's richest families, with receipts — a first chip fab opening at 90nm, frontier models that cost less than a soft-drink launch, and a cricket team that sold for twice everything Indian rocketry has ever raised.

indiaaisemiconductorsresearchopinion

Tracking my mechanistic interpretability self-study

Jul 10, 2026

A running, dated log of working through mechanistic interpretability and AI alignment self-study since late May, mostly following learnmechinterp.com's curriculum.

mechinterpalignment

The Disappearing Parliament: What India's Legislature Actually Does With Its Time

Apr 15, 2026

India's parliament sat for 55 days a year in the 17th Lok Sabha — the lowest of any full-term parliament. 35% of bills passed with under an hour of Lok Sabha debate. In Monsoon 2025, 14 bills became law; the longest LS debate was 34...

indiaparliamentdemocracydatavisualizationpolicy

The Fragmented Calendar: A Data Atlas of Global Stock Exchange Holidays 2025

Apr 10, 2026

Japan loses more trading hours to holidays than any other major exchange. India doesn't have the most holidays — but it has the most fragmented ones. The pattern matters as much as the count.

financemarketsdatavisualizationindiaglobal

Change Detection: Extracting Ground Truth from Noisy Edits

Apr 9, 2026

You want to build an AI system that automatically updates documents when new information arrives — a court ruling updates a legal article, a product release updates documentation. But how do you extract clean ground truth from noisy human edits?

llmevaluationsynthetic-datanlp

The Long Game: The Technical and Strategic Reality of India's PFBR Criticality

Apr 8, 2026

Every PFBR headline just repeated 'Three-Stage Nuclear Power Programme' without explaining what the stages mean or what flows between them. This post is what I found after going looking — with interactive simulations.

nuclearenergyindiaphysics

2025

Fine-Tuning LLMs with Non-Differentiable Human Feedback + RL

May 4, 2025

The blog is focused on giving you an intuition of why even using non-differentiable reward function we are able to use Group Relative Position Optimization (GRPO) for fine-tuning LLMs.

llmrlhfgrporeinforcement learninghuman feedbackfine-tuning

2022

Generalized Visual Language Models

Jun 9, 2022

Processing images to generate text, such as image captioning and visual question-answering, has been studied for years. Traditionally such systems rely on an object detection network as a vision encoder to capture visual features and then produce text via a text decoder. Given a large...

AIvisiondeep-learning