I’m a founding engineer at Lightfield in San Francisco, where we’re building an agentic CRM for companies built with their customers.
The old CRM was designed for the layer between builders and customers. In AI-native companies, that line is starting to blur: sales, product, engineering, delivery, and agents all need the same live picture of the customer. My work sits underneath that shift: public APIs, workflow automation, agent tools, human-in-the-loop review, notifications, and core CRM surfaces.
Before Lightfield, I worked on payments infrastructure at AWS, enterprise workflow automation at Bank of America, graph ML at Capital One, and co-founded Plato, an intelligent service catalog for internal engineering knowledge. Earlier, I worked at Mindgrasp before ChatGPT made AI study tools obvious, and worked on applied ML research at UMD.
The common thread is context: customer context, engineering context, graph context, and the systems that make it usable.
Threads I keep returning to
- context becoming infrastructure
- agents as workflow systems
- customer-facing teams becoming more technical
- internal knowledge becoming queryable
- cities, restaurants, films, and music as taste inputs
Outside work
I keep loose notes on cities, restaurants, films, music, and things I want to remember. Most of that lives in the Log.