Projects
A few representative projects across production agentic systems, robotics, and applied ML.
Multi-Agent LLM Platform for Grid Operations (OATI → CAISO)
Production multi-agent system used to support real-time grid operations workflows, combining retrieval, tool-calling, and evaluation infrastructure.
I helped evolve an early generative-AI prototype into a production multi-agent platform supporting California ISO. The system spans retrieval over hundreds of product documents, tool-calling workflows across internal APIs and data sources, and evaluation/observability tooling to make model behavior measurable and improvable over time.
My contributions focused on building reliable agent behavior (structured outputs, tool selection, and failure handling), strengthening retrieval quality, and adding evaluation harnesses that track tool-call correctness and end-to-end task success.
Adding link to demo soon...
Telepresence Humanoid Robot (Search & Rescue)
Led a 13-person student team building a ROS-based telepresence humanoid robot with perception + autonomy for constrained teleoperation.
I led development across UI, hardware integration, and teleoperation software, coordinating work via Git and an agile workflow. To improve responsiveness, I implemented multi-threaded Dynamixel motor read/write loops and reduced motor response latency significantly. I also built a ROS-based teleoperation system with distributed processes that enabled remote control of the robot’s torso and head using a puppet controller and AR headset.
Images and Video demos being added soon...
Mathematical Reasoning Fine-Tuning (LLM)
Fine-tuned instruction models for improved multi-step mathematical reasoning; focused on evaluation and reliability.
Additional details coming soon...