I am a Member of Technical Staff at Axiom Math, where I work on post-training models for mathematics and software verification.
Before that, I co-founded Alzonova (YC X25), building RL training infrastructure for coding agents that generate formally correct code, and worked on agentic AI for Amazon Q at AWS. I did my MSc in computer science at Oxford, researching multi-agent reinforcement learning and human–AI coordination with Jakob Foerster at FLAIR.
I'm motivated by building personalized, abundant intelligence. Outside of work, I play handball.
-
2026–now
Axiom Math
Core post-training team for models in mathematics and software verification. Designed a new agentic RL training paradigm that became the focus of post-training, and built the evaluation pipeline for test-time scaling and code quality.
-
2025–26
Alzonova (YC X25)
Co-founder & CEO. Built large-scale LLM training infrastructure for RL-based coding agents that generate formally correct code. Backed by Y Combinator, Spring 2025.
-
2024–25
Amazon Web Services
Software engineer on Amazon Q, building agentic AI assistance for the AWS Console and IDE used by millions. Developed the tool-calling and reasoning framework for multi-step tasks.
-
2023–24
Foerster Lab for AI Research, Oxford
Research on zero-shot coordination and ad-hoc human–AI teaming with Jakob Foerster. Implemented multi-agent RL algorithms in JAX.
-
2023–24
University of Oxford
MSc Computer Science, with Distinction. Thesis on multi-agent reinforcement learning at FLAIR.
-
2020–23
FAU Erlangen-Nuremberg
BSc Computer Science, with Distinction. Minor in AI for Biomedical Engineering.
Projects
-
sep 2026benchmarkDrivingBench ↗
Testing frontier vision-language models at driving a real Toyota Corolla around a cone course. With Aditya Ramabadran and Simon Mahns.