• Updated compute infrastructure to Ubuntu 18.04 while maintaining functionality of tools and pipelines for medical image processing and machine learning, transitioning all 12 workstations to LTS and establishing 2 new compute server nodes.

• Conducted instruction through office hours, labs, discussion sessions, online assistance, and grading feedback, allowing students to access and grasp CS 107/107E Computer Systems material easily and efficiently: C, ASM (ARM/x86), systems, sensors, graphics, coding style, etc.

• Managed student repositories to deploy, review, and autograde assignments using Github and Python.

• Developed solutions to further optimize trajectory prediction neural network for FP16 and INT8 inference on Drive AGX platform, resulting in over 2.0x speedup over previously optimized models.

• Implemented workflow to reduce quantization error in INT8 models through quantization-aware training in PyTorch and per tensor calibration in TensorRT (C++).

• Created deep learning workflow from scratch for screening GPUs by capturing frames from traces with DirectX, training networks, and setting performance metrics and benchmarks (Python and C++).

• Developed deep learning solutions on Keras + Tensorflow for supervised and unsupervised anomaly detection in high-resolution frames, validating GPUs on non-deterministic apps with >90% accuracy.