Developed “Erin,” a production-grade autonomous AI agent system for real-time, after-hours lead qualification and follow-ups in a real estate platform. Architected scenario-driven, multi-agent workflows that infer renter preferences (budget, neighborhoods, move-in dates) from SMS interactions and normalize structured outputs for downstream systems. Automated complex workflows including personalized tour list generation and scheduling, resulting in 15+ additional qualified leads per day and a measurable increase in sales conversion and agent productivity.

As a founding member of the BuckyBots team, I led the software development for the Nao robot platform from the ground up. To overcome the data scarcity and time constraints of physical robot training, I engineered a high-speed simulation environment using PyGame. This Digital Twin approach enabled data acquisition at 500 samples per second—roughly 2,500 times faster than physical hardware—accelerating the development of robust reinforcementSim2Realreinforcement learning models.

My technical contributions spanned the entire robotics stack: utilizing OpenCV for real-time perception of balls and field markers, implementing table-Baselines3 for behavioral planning, and optimizing robot gait and control in C++based on the B-Human framework. These efforts culminated in securing 3rd place in the Standard Platform League at RoboCup 2023 in France.

Explored the integration of hardware and cloud software through a custom IoT monitoring system. I performed precision soldering and hardware integration for the BME280 sensor via I²C communication on a Raspberry Pi. The system features a Python Flask backend that processes environmental data in the cloud, providing a real-time web dashboard for monitoring room temperature and humidity alongside personal schedule management. This project demonstrates my ability to bridge the gap between physical sensors and web-based data visualization.