Christopher Babayan I am a physics undergraduate at the University of California, Berkeley, with experience in research, machine learning engineering, and developing automated task-generation pipelines and evaluation harnesses.
Research Intern
Led the design and development of automated task-generation pipelines producing verifiable RL environments for frontier-model post-training across three domains: enterprise computer use, software engineering, and knowledge work.
Technical Consultant
Machine Learning Engineer (Consultant)
Integrated an MXFP8 quantization path into TorchAO via a custom tensor subclass with overrides for matmul, linear, and transpose operations. Developed custom NKI kernels for PyTorch quantization operations, including a fused MXFP8 matmul kernel for LLM inference on Trainium that
Undergraduate Researcher
Built a differentiable PyTorch data-generation pipeline producing neutrino crossings from flux and energy moments via Minerbo maximum-entropy closure.
B.A. Physics
