Experience

Embedded Linux Smart Camera

• Delivered core features and production hardening for a C++17 embedded Linux camera firmware targeting a 50,000-unit production run. Designed auto-white-balance and adaptive low-bandwidth streaming, and led migration of the cloud backend to AWS IoT; hardened the firmware across hundreds of commits covering OTA recovery, image-pipeline reliability, and state-machine edge cases.
• Led development of the manufacturing test fixture — a multi-threaded Python + Arduino station that programs and tests up to 8 STM32MP157F devices in parallel via USB DFU and SSH, SQLite manufacturing history, and a two-phase sealed-device workflow covering OTP secure-boot programming, certificate provisioning, and OTA production-firmware install.
• Brought up a custom Yocto embedded Linux distribution for STM32MP157F and architected end-to-end secure boot pipeline, from TF-A (mbedtls X.509 verification) → U-Boot → FIT-packaged kernel (verified via a custom two-level master/leaf-key hierarchy) → DM-Verity-protected rootfs for A/B Bank OTA with anti-rollback; fully post build signing using AWS CodeBuild+KMS+Secrets Manager.

Autonomous Construction Robot

• Designed a CANopen supervisor MCU targeting STM32H735 running C++17 to communicate with 4 motor controllers and a ROS2 PC. Reverse-engineered the stock CANopen protocol from candump traces, PDO configuration dumps and vendor docs and implemented a state machine architecture to track and send heartbeats, control commands, EMCY alerts, etc.
• Designed an RS-422 quadrature encoder interface PCB in KiCad using AI tools, and brought it up with RP2040 firmware that decodes 4 encoders in PIO and streams over RS-485 at 400 frames/s with zero drops. Also brought up the robot’s RS-485 Modbus IMUs and CANopen inclinometer on the bench.

Smart Batting Glove

• Rapid prototyped proof-of-concept hardware on Arduino Nano including bringing up a shared I2C bus for 10 capacitive force sensors, firmware to buffer data, detect swings, and stream data over BLE to a local host web interface.
• Brought up the production firmware stack on nRF54L15 / Zephyr by writing IMU drivers for both I2C and SPI. Added GitHub Actions CI with native_sim tests. Used AI tools to rapidly build a React Native companion app to stream data via BLE GATT from device to mobile phone.

• Integrated a dual-redundant EPAS steering rack onto the HIL validation rig and extended the Pytest suite, bringing 300+ nightly tests online to verify the rack’s CAN responses against vehicle-level stimulus.
• Built a fleet-scale alert dashboard across 1M+ vehicles that the team used to judge each firmware release (SQL, 10× query speedup). Diagnosed anomalies with CANalyzer traces and PySpark notebooks and handed findings to firmware engineers.