I am a Lead ML Research Engineer at the Kempner Institute at Harvard, specializing in performance and efficiency in machine learning. My current projects center on large-scale training and inference systems, including KempnerForge, an open-source PyTorch-native framework for fault-tolerant distributed training of foundation models that I help develop, and my MLPerf Inference v6.1 submission of Llama 3.1-8B in FP8 on a single NVIDIA H200. My work spans large-scale experiments on diffusion models and transformers, performance benchmarking on GPU clusters, and the development of reproducible open-source codebases. I focus on rapidly implementing research ideas, profiling and optimizing bottlenecks, and delivering systems that make AI experiments faster, scalable, and reliable.

Schedule a Meeting

I’m open to collaboration, research discussions, and new ideas. I’m always happy to connect with fellow researchers, students, and industry professionals. These sessions are part of my commitment to the community and open to all.