Health Technology - Real-Time Audio & ML Platform

• Architect for hearing-mode switching on Ray-Ban Meta AR glasses; defined cross-stack design across

firmware, on-device middleware, and mobile companion UI spanning 4 engineering teams.

• Led bring-up of Conversation Focus speech-enhancement feature across multiple wearable device

generations; drove cross-stack debugging from firmware to runtime.

• Conducted 60+ System Design interviews (certified interviewer); mentored 8+ engineers and research

interns as onboarding buddy.

• Delivered 5+ internal tech talks on developer tooling, debugging, and ML platforms.

• Top 3 in org-wide Engineering Excellence Fixathon (1st in division).

Audio Research @ Reality Labs

• Rearchitected the core thread-scheduling subsystem powering shipped real-time audio pipelines,

cutting end-to-end latency 65% (300ms → 106ms); parallel build-pipeline overhaul reduced deployment

time 3.3× and binary size 15×.

• Refactored core real-time signal-processing modules (VAD, post-filters) and integrated sensor-fused

(IMU-based DOA) beamforming into the production audio pipeline.

• Architected an edge ML deployment platform (PyTorch → libtorch on NVIDIA Jetson) with versioning, hot-swap, FP16 inference, and pytorch-vs-libtorch parity testing; adopted by 6 research scientists to iterate on 80+ ML models in real time.

CoreMedia Network Playback

• Led error-reduction effort on the cross-platform C/C++ networking and playback stack (iOS, macOS, tvOS, watchOS); eliminated 4 of the top 10 errors and added observability on the remainder.

• Owned the on-device telemetry library for playback and download performance metrics, deployed across all Apple devices worldwide; drove ML-based performance investigations.

Part-time Research Assistant @ Image and Vision Processing Lab under supervision of Zhou Wang

Android Platform

• Designed test automation for bootloader fastboot mode and recovery kernel on NVIDIA SHIELD (C); reduced test turnaround time 20\%.