• Systems engineer building OS-level audio and on-device ML infrastructure in modern C++ under strict real-time, power, and latency constraints.
• Led system architecture for an always-on, DSP-accelerated audio inference capability on custom OS hardware.
• Designed low-latency streaming audio pipelines enabling partial-window inference and continuous audio capture across detection boundaries.
• Built and maintained a native audio input service on AOSP (Android Open Source Project), focusing on deterministic, real-time capture and shared-memory IPC.
• Contributed to the evolution of the core audio stack from Android (AOSP) to embedded Linux, adapting services and execution models accordingly.
• Optimized on-device inference workloads via model quantization and DSP offloading to hardware accelerators.