Collaborating with a team of 20, I developed an agentic AI assistant for IBM's ETL tool, Datastage. I implemented over a dozen feature capabilities to be deployed in production environments used by most state governments and many international banking agencies. Each feature relies on natural language parsing through many-shot prompting and integration with IBM's internal REST APIs. I also led a redesign of the conversation framework to improve back-and-forth flow and human-in-the-loop communication.

I spent a semester as part of Professor Andrea Bajcsy's Intent Lab, dedicated to developing safer reinforcement learning policies for robots through world models and reachability analysis. I provided feedback on upcoming papers about mitigating the impact of distractors when using world model predicted rollouts by using VLMs to remove irrelevant sections of an observation.

I was a teaching assistant for CMU's largest CS course, 15-122. The ~500 students who take the course each semester build general DSA knowledge and learn C with all its memory-level complexities. My job as a TA was to help design course content to encourage conceptual understanding and retention, reinforce key concepts learned in lecture through small-group collaboration sessions, and target fundamental misunderstandings in office hours, setting them up for many higher-level CS and ECE courses.