Mem
Engineering Manager and Senior Product Engineer
I managed six engineers building Mem's AI assistant and led the engineering team while our CEO was on leave. I stayed hands-on the whole time, building alongside the team. When we rebuilt Mem from the ground up as Mem 2.0, I developed the new frontend and contributed to the design system behind it.
- Led the team through the shift to agentic development, continually evaluating new coding agents and making them central to how we built AI features. I used them for all of my own development work over the last year.
- Led Content Understanding across in-note OCR, the frontend, agent context, and a backend content pipeline, helping the assistant see into note attachments. Usage showed people cared less about huge PDFs and more about the figures and attachments in their notes, with questions like “Is this signed?”, so that's what we shipped first.
- Changed how we shipped. On Content Understanding, I introduced earlier technical planning and phased releases, and they went out with virtually no post-release bug reports, compared with weeks of fixes after earlier launches.
- Built connected services end to end: evaluated vendors and chose Composio for third-party integrations, then built everything from authentication and permissions through data retrieval and background processing, including the setup and recovery flows customers actually see.
- Built assistant features around Mem's custom Python agent, including proactive playbooks that turn connected data into tasks, third-party connectors, and iMessage and WhatsApp channels.
- Worked with the leadership team to plan an AI executive assistant that spots commitments across your tools and organizes follow-ups.
- Wrote agent evaluations for source accuracy, task ownership, duplicate prevention, and notification relevance, so the assistant's suggestions were worth paying attention to.
- Shaped our data ingestion and agent reasoning pipelines through design and review: resumable processing, deduplication, and traceability back to original sources.
- Helped build production-support agents that trace issues, post Datadog evidence in Slack, and propose fixes they can build, test, and deploy. Resolution time went from hours to minutes.
- Led our SOC 2 implementation, coordinating engineering work and security controls.