Experience & education

Cruise — General Motors

AI/ML software engineering intern

Vehicle fleet motion reconstruction using ML

• Built an ML system to reconstruct missing vehicle motion in recorded-driving simulations; reduced best-candidate average position error by 24% versus the team’s prior baseline on approximately 58,000 held-out vehicle trajectories.
• Built a scalable data pipeline for 38,000+ highway scenes, processing terabytes (TBs) of raw tracking data with Google Cloud Storage and Dataflow for scene reconstruction.
• Built a large-scale training pipeline across multiple H100 graphics processing units (GPUs) with PyTorch DistributedDataParallel (DDP), the NVIDIA Collective Communications Library (NCCL), and mixed precision, reducing wall-clock training time from 12 to 3 hours.

Mitsubishi Electric Research Laboratories (MERL)

ML and optimization research intern

• Developed a physics-constrained model combining neural ordinary differential equations (Neural ODEs) and gated recurrent units (GRUs) for vapor-compression heating, ventilation, and air conditioning (HVAC) systems, improving forecasting accuracy by 14% and achieving a 5.7× runtime speedup versus the prior modeling approach.
• Delivered reusable PyTorch software for data processing, training, evaluation, and GPU-accelerated control experiments with limited real-world data.