Resume

moved to the west coast 🌁

I develop deep learning models to simulate and optimize semiconductor manufacturing processes.

• Debugged PyTorch implementations of distributed training and inference systems, including GPipe, 1F1B pipeline parallelism, vLLM inference, and post-training quantization.
• Evaluated student projects on agentic AI, ReAct, Model Context Protocol, and Direct Preference Optimization for Qwen and FLAN-T5 models, assessing implementation correctness, evaluation quality, and reproducibility.
• Held weekly office hours for 110 graduate students, working through training instability, memory and throughput issues, and implementation errors.

• Developed a Hierarchical Neural Network in PyTorch to model Parkinson’s disease progression from clinical data.
• Improved training efficiency by 33% over a standard neural network baseline using V-cycle and W-cycle multigrid methods.

• Identified key barriers to maternal health care in Nepal’s Terai region as part of a three-member team.
• Conducted interviews and focus groups with mothers, healthcare workers, and local government officials, presenting findings to senior leadership.

• Employed a genetic suppressor screen to identify the genes involved in regulating phototolerance in Arabidopsis thaliana during iron deficiency.
• Conducted downstream molecular experiments to measure gene expression, presenting the findings at the Karen E. Wetterhahn Science Symposium.