RSSAmplifier

Blog

Shubham's blog

Hello there, welcome to my blog! I write mainly about my exploration into AI/ML research and engineering. Took it up to stay up-to-date with the most recent ...

shubhamg.bearblog.devRSS feed ↗10 posts

Latest posts

Kazakhstan- Notes from a Country I Knew Nothing About

Seven days in Almaty I. The Choosing We did not choose Kazakhstan. Kazakhstan was what remained after everything else had been eliminated, which is a different thing entirely, and I say this without embarrassment, because most of the best places I have been to arrived in my life the same way. What we knew was what we did not want. We did not want another beach. Between the three of us we had done…

Sri Lanka trip -The Safest Place to Be Afraid

Six days in the water, and one long drive inland I Colombo is not far. The map says so plainly — a short hop across a narrow stretch of sea, nearer to me than half of my own country. It was the connecting flights that made it feel like a journey, that stretched a hundred-odd kilometres of ocean into the better part of a day. Then I landed, and the arrivals hall was playing Bollywood. It was both…

"It's Not a Lie If You Believe It": LLMs Defend Their Most Fluent Memory Against Everything — Including You Being Right

=1200px) it renders as an always-visible sidebar in the left margin, Anthropic-style. The checkbox and hamburger button are hidden and unused. - On narrower screens it collapses behind a floating button in the bottom-right corner; the hidden checkbox holds the open/closed state and a tap anywhere closes it. The links point at the anchors already sitting above each heading in the post, so they work…

Claude improvement ideas

A list of ideas that would benefit Claude (and some other chat providers as well). I will keep adding to this list as and when I think of new ideas! Ability to fork chats - What if I could choose parts of the chat I want to share and allow people to play with it? Just like you can fork code in GitHub, there should be an option to fork chats in Claude. Within a chat, you can select the responses…

Indic NLP research: AI4Bharat papers - Part 2

Building on the Foundation: IndicBART and IndicNLG Suite This is a continuation of the AI4Bharat papers series. If you haven't read Part 1 , I highly recommend starting there, as each paper builds upon the previous work. The AI4Bharat Journey: Where Part 2 Fits In In Part 1, we explored two foundational papers from AI4Bharat: IndicNLPSuite (Nov 2020) : Created IndicCorp (8.8B tokens), IndicBERT…

Notes on RLHF | Nathan Lambert's book

Overview Nathan Lambert is a distinguished researcher from AI2 (Allen Institute for Science) who recently published a book on RLHF (Reinforcement Learning from Human Feedback). Book can be found here Core Concept RLHF enables a base model to learn from human preferences. These preferences are encoded in a separate reward model, and the training objective is to align the base model with human…

Indic NLP research: AI4Bharat papers — Part 1

The series aims to delve deeper into each paper and understand what's being done in Indic language space. One of the pioneers in this has been the research institution 'AI4Bharat'. An overview of how the NLP research done by AI4Bharat from 2020 onwards builds on top of another ( image credits: Nano banana pro from Gemini ) This is an attempt at exploring the first 2 papers that came out in the…

Conversational Intelligence Platform for Indian BFSI

I had done this analysis as part of a hiring assignment looking to enter the Conversational Intelligence market. Table of Contents Market Research Market Landscape The Opportunity Problem Statement & Hypothesis Target Audience & Customer Segments Product Vision & Strategy Product Vision Value Proposition Success Metrics MVP Definition & Design MVP Core Features UX Design Go-to-Market…

Owning your data - A model for everyone

Imagine if each person was in control of all the data that they were producing as digital trails. What is it that could tell us about ourselves? Could I use it to do something more useful than being suggested books from Amazon? Today my data is siloed — my want-to-read books are in Goodreads; my purchases are on Amazon; and many of my want-to-read books are sitting in my bookmarks folder. Suppose…

My NLP/AI learning journey

A personal documentation of diving deep into AI - from NLP foundations to speech AI and beyond Why I took this up The idea behind learning AI was mainly to stay at the cutting edge of technology. The focus and development of LLMs at OpenAI has a fascinating beginning; it wasn't planned but emerged from experiments, and the early team took note and decided to go all in on it, thereby transforming…