I work across the full machine learning lifecycle, from designing new models to building production-grade pipelines for enterprise AI systems. My recent work spans large-scale model training, multimodal learning, agentic LLM systems, and safety-critical AI.

My research has focused on machine unlearning, decentralized LLM pre-training, privacy-preserving learning, and AI safety, with publications in venues including NeurIPS, ICLR, and SIGMOD and funded work on privacy-preserving and safety-critical machine learning systems.