About me

Hi, I’m Adib Hasan. I build long-horizon agents that improve through expert iteration in context space, using rich feedback from verifiers, tools, and real-world execution.

Using this approach, my agents have discovered 150+ vulnerabilities in widely used OSS projects, including Next.js, pnpm, and MetaMask; improved an upper bound for an open math problem; and topped the Spider 2.0 DBT benchmark for data science tasks. Previously, I spent two years as a quant researcher managing multi-million dollar crypto portfolios and leading a team of four. You can find more on my LinkedIn.

I completed my bachelor’s in Math & EECS and MEng in EECS at MIT, where I was advised by Dr. Mardavij Roozbehani and Prof. Munther Dahleh. My research focused on model robustness under distribution shift, resulting in two papers: a variational pretraining framework for weather time series (AAAI 2026 Oral), and a pruning-based jailbreak defense for LLMs (EMNLP Workshop 2024).

I grew up in Bangladesh, where I spent much of high school training for and participating in the International Mathematical Olympiad. I also compiled a collection of my olympiad preparation notes from those years here.

Beyond work, I cook passionately to avoid starvation (proof), read dead philosophers, and lift heavy objects calmly yet furiously. I also co-authored two math textbooks and built a free college admissions GPT. Both make high school students cry, but for completely different reasons.