Haixi Zhang

Hi, I’m Haixi Zhang, an ML / Computer Vision / Robotics Engineer based in San Jose, CA. I hold an M.Eng. in Electrical & Computer Engineering from Cornell University and a B.S. in Electrical & Computer Engineering from the University of Rochester.

I build production perception systems that combine multimodal sensor data, 3D geometry, and deep learning. At Tensor Auto, I am the sole technical owner of three ML/CV systems, from network design and training to data pipelines and onboard inference. I designed, trained, and deployed two neural sensor-calibration systems. I also built and evaluated front- and rear-seat occupant perception, combining person detection, posture classification, iToF reasoning, and VLM temporal fusion.

FEATURED INDUSTRY WORK Neural Perception Systems—Designed, Trained & Deployed Camera–LiDAR and LiDAR–LiDAR calibration built from network design through TensorRT/C++ deployment, alongside integrated occupant perception. 3 systems owned end to end 2 production deployments 95% less rotation error · unseen vehicle 63% less fleet rotation error · LiLiNet LCCNet · Deployed LiLiNet · Deployed In-Cabin Perception · Vehicle Evaluated Explore selected work →

Technical Focus

Neural Model Design & Geometric Learning

Designing dual-branch encoders, cross-modal attention, camera-conditioned FiLM, and spherical correlation, with SE(3) / SO(3) objectives for sensor calibration and multi-task visual learning for occupant perception.

Production ML Engineering

Engineering world-size-configurable PyTorch DDP / NCCL training runtimes, then carrying models through ONNX / TensorRT optimization, performance profiling, and production C++ integration.

Data & Evaluation Systems

Built a shared Airflow/Kubernetes sensor-data pipeline that streamed 17+ TB of raw logs in a single pass to produce ~600 GB of calibration data, with measured clock synchronization, traceable labels, and view-aware selection.

I enjoy working across the full perception lifecycle—from data and modeling to runtime integration and benchmarking—and collaborating with cross-functional teams to improve end-to-end system quality.

Outside engineering, I’m fascinated by world history from the Renaissance through the Second Industrial Revolution, and I remain an enthusiastic fan of Yes, Prime Minister.

Interested in production perception, autonomous systems, or applied computer vision? I’d love to connect.