Aug 2024 – Present · 1 yr 7 mos · Zurich, Switzerland

• Developed and optimized post-training recipes for the Llama 4 family of models, enhancing model performance and alignment through advanced fine-tuning techniques
• Spearheaded implementation of comprehensive benchmarking suite including GAIA2 and scalable evaluation pipelines, establishing robust frameworks for agent capabilities assessment and creating reproducible recipes for systematic performance measurement
• Architected and built multi-agent research infrastructure with complex simulated environments, enabling sophisticated agent interactions, automated annotations, and novel benchmarking methodology development
• Engineered end-to-end data pipelines supporting pre-training, reinforcement learning, and agent infrastructure, ensuring efficient data flow and processing at scale
• Pioneered development of specialized AI agents including deep-research, code generation, and strategic planning agents, advancing autonomous reasoning and task execution capabilities

Apr 2022 – Jul 2024 · 2 yrs 3 mos · Zurich, Switzerland

• Designed and implemented end-to-end camera, LED, and IMU calibration solutions for Meta's Quest products, driving improvements in measurement accuracy and device reliability
• Led cross-functional teams in creating scalable calibration frameworks for multiple headset and controller models, significantly reducing reprojection error and advancing key performance indicators
• Owned and developed robust infrastructure for both factory and online calibration, enhancing quality assurance processes and enabling continuous post-shipment performance monitoring