AI researcher and engineer with 18+ YOE in multimodal LLMs, machine learning, physical-world AI, and computational modeling. Previously Principal Machine Learning Engineer at Roblox, developing 3D vision-language models for content understanding and safety, and Software Engineer, Machine Learning on Meta’s Visual-LLaMA team, developing multimodal foundation models for Ray-Ban smart glasses. Research expertise spans vision-language learning, visual grounding, hallucination reduction, and large-scale model post-training.
Optical and hardware engineer with 10+ YOE specializing in high-speed data-center interconnects and optical communication systems. Advises PhyNexa on hardware architecture and interconnect design for efficient GPU/CPU computing infrastructure.
Software Architect & Full-Stack Engineer · Consulting Partner, Pan Digital Labs
Former Software Engineer at Amazon, specializing in software architecture, full-stack development, and scalable systems. Advises PhyNexa on API design and platform engineering.
Research Collaborator — collaborating on failure modes of VLMs and the Gated LoRA algorithm for lower memory.
XM
Xiaofu Ma
PhD, Electrical and Computer Engineering, Virginia Tech
Research Collaborator — collaborating on VLM modeling of crystal fracture of objects under contact, fast graph search algorithms for classification, and the Hard Negative Visual Encoder loss for low latency and memory.