Work

2025–present · San Francisco

I work across the product and systems behind Lightfield's CRM. I've helped turn product operations into a public API and Python SDK used by our own agent, built tools for creating and editing CRM tasks, and worked on workflow automation, human review, notifications, and core product surfaces. The recurring challenge is translating an ambiguous request into a change that is understandable to the user and correct in the product.

2025

I joined YouTube's Living Room team and left after a week to join Lightfield. I wanted more ownership and a shorter distance between a product decision and the customer affected by it. It was also a high-variance bet made with limited information; that was part of the appeal.

2024–2025

We built an intelligent service catalog for internal engineering knowledge: natural-language service search, organized documentation spaces, and a support agent grounded in team documentation. We applied to Y Combinator but did not turn Plato into a lasting company. It was my first attempt to choose the problem, build the product, and convince other people it should exist.

2024 · Seattle

I built reporting and reconciliation infrastructure for roughly ten million financial events a month. The system used SNS, SQS, Lambda, DynamoDB, EventBridge, S3, and CloudWatch to trace events across services and preserve reporting completeness through retries and partial failures. The point was not throughput by itself. Financial data had to remain trustworthy when the underlying delivery system was asynchronous and imperfect.

2023 · College Park

I worked with a roughly 900-million-edge graph representing relationships in card data. I rewrote motif queries with GraphFrames and improved their performance by about six times. The project taught me that a graph is useful only when its relationships can answer an operational question quickly enough to affect the product; scale alone is not the interesting part.

2023 · Jersey City

I automated part of an internal risk-testing workflow with Python and SQL, reducing its runtime by roughly 85 percent. The work was less glamorous than the number suggests: understand how analysts performed the process, preserve the checks they trusted, and remove the repetition without making the result harder to inspect.

2022 · College Park

I joined Mindgrasp before the product had traction and worked directly with the founders from their first UMD office. We were applying OCR and language models to lectures and study materials before most students had used a conversational AI product. It was my first close look at models becoming a product rather than remaining a demo or research result.