Recent Writing
Understanding LLM.int8(): Making Large Models Fit
Fri Jun 12 2026
Understanding Flash Attention: Making Transformers Fast
Tue Jun 09 2026
Transformer Architecture: From First Principles
Mon Jun 08 2026
Ankit Mishra
I build software and A.I products from 0 -> 1, move fast, and thrive in high-ownership environments where initiative matters.
Founding AI Engineer at DocuraHealth (YC W26), ex-MLE at Pibit.ai (YC W21), and ex-Founding AI Engineer at AarogyaID . During college, I worked across 5 startups and an open-source organization as an AI & Software Engineer and also taught as a Python Instructor at USF.
Building agentic software that creates real-world value.
THE MATH THAT BENT THE TRAJECTORY OF AI
Attention(Q, K, V) = softmax(QKT/√dk)V
Lex Fridman — Go All The Way (Charles Bukowski)
0:00
Latest Update
Launched mutant-ai v0.7.3
An adversarial generation pipeline for LLMs. Turn standard evaluations into behavioral stress tests to uncover hidden failure modes.
[ view on pypi ]
> pip install mutant-ai