- π I've been working on AI for code (nΓ©e ML4Code) since 2012. I research and engineer AI systems to make developing and maintaining software productive, enjoyable, and less error-prone.
- π€ As of June 2026, I am at Apple. Previously I was a staff research scientist at Google DeepMind and a principal researcher at Microsoft Research.
- π I enjoy building AI, practical developer tools, training and orchestrating agentic systems, and solving real-world problems with machine learning and optimization methods.
π€ LLMs & Agentic Systems
- π Evaluation of LLMs and agentic systems
- π΅ Agentic systems for cybersecurity
- π Post-training: RL for software engineering tasks
- β‘ Architecture of ML Models
π» AI for Software Engineering
- π Automated bug detection and program analysis
- β¨ Code generation, synthesis, and analysis
- π¦Ύ AIβhuman collaboration in software engineering
- π¨ Practical developer tools
- π The interplay of formal, sound analyses with LLMs and agents
π₯ Highlighted Publications (π Full list)
As the industry has moved away from publishing, I have also ceased to work/submit academic publications.