Research Engineer
Atul Saxena
I build the environments models are trained in, and the tools that show what they learned.
- RL Systems
- Interpretability
- ML Infrastructure
Research Engineer
Atul Saxena
I build the environments models are trained in, and the tools that show what they learned.
Research Engineer — RL Systems · Interpretability · ML Infrastructure
I build the environments models are trained in, and the tools that show what they learned.
3
Published papers
10
Merged upstream PRs
8
Projects contributed to
3
Packages on PyPI
I work on reinforcement-learning environments, mechanistic interpretability, and the infrastructure that holds both together.
Right now that means building production RL environments at Snorkel AI — task design, reward and verifier logic, and figuring out whether a failure came from the model, the environment, the grader, or the spec. Before that I led a pod at Bespoke Labs on agent evaluation and data curation for frontier-model work.
The rest of the time I run Blackdrome AI Labs, where we publish what we find. Three papers so far: a mechanistic-interpretability study of how GPT-2 routes semantics, a neuro-symbolic maths engine in Rust, and an open decision model with released weights.
I spend a lot of time in other people’s repositories. Kornia, Pandera, Qdrant, toqito, OpenMed, Monocle, Bespoke Curator. Usually something small enough to understand and annoying enough that I couldn’t leave it alone.
I care about whether an experiment survives its controls, whether a benchmark measures what it claims to, and whether the code still works when the environment gets weird.
Indian Institute of Technology Madras
B.S. in Data Science and Applications
Chennai, India · Ongoing
Techkriti Hackathon Finalist
IIT Kanpur
Cyber Sentinel Winner
IIT Madras
Top 250, Shipped Pre-Accelerator
Lovable
Live industry research project for HMC Group (Hero MotoCorp)
Business.io, IIT Madras · Jul 2025
Professional Diploma in Agile & Project Management — MTF Institute, Portugal