📍 Geneva, Switzerland | 🔬 Systems researcher and entrepreneur | 🎓 EPFL → MIT → MPI-SWS

I lead the Data Systems Group at MPI-SWS. We work across systems, databases, and machine learning, and most of what we build now is infrastructure for agentic AI: systems that coordinate many models over data far larger than any one of them.

• 🛡️ scode – Sandbox wrapper that keeps AI agents away from your credentials
• 🎛️ codemux – One CLI for thirteen AI coding agents, every run sandboxed by scode
• 📊 usagemux – Subscription quota and credits, normalized across providers

See also: mpi-dsg for more research prototypes.

• 🌊 X-Stream – Edge-centric graph processing using streaming partitions
• 🌀 Chaos – Scale-out graph processing from secondary storage
• ⛈️ Hailstorm – Disaggregated compute and storage for distributed LSM-based databases
• 🌪️ Hurricane – Taming skew in large-scale analytics
• 🔮 Tesseract – Tesseract implementation in Python
• 🗃️ Loki – Generate databases from high-level workload statistics

• Agentic AI Infrastructure – Building systems that let AI agents operate at scale
• Cache Saver – LLM inference optimization framework
• Code Intelligence – SQL-of-Thought and LoRACode, teaching models to reason about code
• LLM-Enhanced Databases – Rethinking query processing with learned components
• Teaching – Operating Systems, Distributed Systems, and Systems for LLMs at Saarland

"Understand the system before you optimize it."

I build data systems that scale and teach machines to reason about code. The best research ships as real software.

• I have moved between research and startups my whole career
• PhD at EPFL, postdoc at MIT CSAIL, now group leader at MPI-SWS
• Swiss Armed Forces veteran (IT Pioneer / Chief of Staff)
• I work on systems that end up running in production, not only in papers