Iterating on model architectures, sensor inputs, and data quality.

Measuring performance, locating failure modes, and burning down problems.

Collaborated to extend sensor fusion models to longer ranges resulting in better long range perception and fewer behavioral issues on freeways.

Conducted latency analysis and tuned computational graphs to increase onboard throughput and unblock model launches.

Rapidly iterated on ML data extraction and sensor synchronization to facilitate sensor fusion experiments and reduced data distribution shift.

Unified ad hoc data mining pipelines into a standardized mining framework to pay down technical debt and facilitate easier data drill-down.

Built a novel labeling workflow and gathered a dataset of localized long range camera labels to improve camera localization models and validate existing evaluation methods.

Teaching Assistant for the "Foundations of Computing" course at the University of Washington covering logic, proofs, set theory, and other discrete mathematics. Responsible for teaching 30 students in a weekly section on fundamentals of CS theory, hosting weekly office hours for questions on material/homework, and grading weekly assignments.

Implemented a large scale data pipeline in C++ to publish weekly digital rights management (DRM) reports for YouTube's partners. Also assisted in migrating DRM APIs from Python to C++.