Worked on neural decoding and application-level software for the N1 implant, supporting real-time intent prediction and adaptive feedback in user interaction tasks such as cursor control and text entry.

Helped improve interaction accuracy and usability through iterative tuning of decoding parameters and closed-loop visual feedback, informed by patient sessions and clinical feedback.

Built and maintained high-bandwidth neural data pipelines handling thousands of channels with tight latency constraints, following FDA and IEC 62304 requirements for medical software.

Collaborated closely with clinicians, researchers, and patients to refine calibration workflows and improve long-term reliability of BCI applications.

Developed and maintained real-time multi-sensor fusion pipelines combining lidar, radar, and camera data for urban autonomous driving.

Implemented object tracking and sensor fusion algorithms, including Kalman-filter-based approaches, to improve robustness under sensor noise, occlusion, and adverse conditions.

Worked on performance tuning and latency reduction across the perception stack to support large-scale on-road testing and safety validation.

Partnered with mapping, planning, and safety teams to ensure perception outputs were reliable and usable in production driving systems.

Worked on performance-critical C++ and CUDA components for real-time graphics and physics systems, with a focus on latency reduction and predictable execution.

Applied cache-aware memory layouts, low-level optimization techniques, and hardware-software co-design to improve throughput and determinism.

Contributed to internal GPU compute platforms used across multiple product generations.