Wearables Contextual AI team

• Built the MVP hardware, firmware, and enclosure for a camera-based, battery-powered, LTE-connected smart meter reader on a tight 1 month timeline.

• Acquired 2 paying customers for our smart meter reader, and built two ruggedized and weather-proof versions of our device to deploy on-site at their buildings.

• Upgraded to V2 magnetoresistive smart meter readers for our existing customers to give them a 15x improvement in time granularity and a 500x improvement in measurement precision.

• Authored the backend software in Python and Django.

• Visualized our deployed IoT devices' metrics on our internal dashboard using Grafana.

• Created the MVP of our building energy data analytics solution using Python, Pandas, Streamlit, Plotly, and Firebase.

• Launched the MVP of our SaaS tool to 40 beta users. Conducted user interviews with them to gain feedback, and released major feature launches every week.

• Interviewed, hired, and managed a design consultant to develop our outreach materials.

• Composed the communications logic between Schneider Electric's solar inverters and the smart grid protocol IEEE 2030.5, using embedded C.

• Tested the system for 60+ checks of correctness using Python, Gherkin and the BDD methodology.

• Architected state-based design in C to integrate NFC and Bluetooth (BLE) on IoT smart heated gloves

• Composed the OTA (over the air update) feature to update the firmware via BLE from a smartphone

• Wrote Python scripts to automate the build process and device flashing process for manufacturing

• Transferred PyTorch computer vision algorithms from CPU to GPU, and used multi-threading, to improve the video inference engine's speed by 2x.

• Identified a critical memory leak that caused the video inference engine to crash every two hours. Fixed the leak to reduce memory usage to 12GB RAM - a 5x reduction from before.

• Profiled the AI video pipeline using the Python profiling tools Memray and PyInstrument.