Engineering leader focused on AI tooling, platform infrastructure, and scaling teams that ship.
Member of Technical Staff at Arcade.dev, leading the Tools, Growth, and Discover & Build teams. Arcade is building the platform that lets AI agents actually do things: tackling auth, permissions, security, and connecting to the systems we've spent decades building. I lead the teams responsible for what those tools are, and how developers discover, build, and adopt them.
At Arcade, I'm focused on the infrastructure that makes AI agents useful in the real world. Not just chat - actual tool use with real auth, real permissions, and real security. Tools builds what agents can do. Growth owns our PLG/PLS motion and experimentation. Discover & Build helps developers find and build agents with Arcade.
Previously at A Place for Mom, I led the team behind Grace, a production multi-agent system powering an AI-native senior care marketplace. Key architectural decisions: hexagonal architecture for LLM provider flexibility, composable prompt packs with filesystem fallback, and a dual-layer guardrail system combining pattern matching with LLM evaluation.
MCP server that takes a product requirements doc and decomposes it into Jira-ready epics and stories. Built with OpenAI Agents SDK. The part I'm most interested in is the eval suite: how do you measure whether a PRD was "correctly" decomposed? That's a harder problem than the decomposition itself.
5+ years of engineering leadership, Manager → Sr Manager → Director across MarTech, Platform, and AI teams. Hired and developed 15+ engineers from IC to Principal level.
Arcade.dev — Member of Technical Staff. Joined as the team's first Engineering Manager, now leading Tools, Growth, and Discover & Build. Building the platform that connects AI agents to real-world systems with proper auth, permissions, and security. We raised a $60M Series A this year, bringing total funding to $72M.
A Place for Mom — Director of Software Engineering. Led 12 engineers across AI Product and B2B Platform. Presented product strategy to CEO and executive leadership. Transformed a 21-person platform org: KTLO from 45% to 19%, P0 incidents from ~1/week to 1/year, delivery lead time from 20+ days to 6 days. Rebuilt the leadership team across an 18-month transformation: hired a QA Manager, Sr Manager, 2 Principals, 2 Staff Engineers, and promoted from within.
Realtor.com — Manager → Sr Manager across three teams that grew to 18 engineers (Auth, Identity, Notifications, MarTech). Founded the MarTech org from zero. Scaled Notifications from a single-use system to a company-wide platform doing 4B+ quarterly. Led Auth0 migration of 170M user records with zero downtime. Partnered with Data Science to productionalize ML-powered recommendation models.
Pandora — Part of a 4-engineer founding team that built the company's distributed message bus from scratch on Kafka. It became the foundation for all future data platform work.
Lightspeed Venture Partners — Founded a VC-backed startup through their fellowship. Didn't make it, but learned when to build vs. buy and how to validate before you've built anything.
Okta — SDR → AE Emerging → AE Corporate. Deliberately left engineering for enterprise sales to understand how customers evaluate, buy, and adopt platforms. Presidents Club 2019. Came back to engineering. It changed how I build - for adoption, not just technical elegance.
Technical depth: AI/ML infrastructure (multi-agent systems, agentic AI, reinforcement learning), distributed systems (Kafka, microservices), platform engineering (Auth, Identity, Notifications, Search). Computer Engineering, Santa Clara University (Dean's Scholar).
How to make AI agents actually useful beyond chat: the auth, permissions, and security problems nobody wants to solve. Whether agents should find their tools at runtime instead of loading everything upfront, and what that means for tool design - that's the bet behind Omni. How you evaluate a skill or a tool when the market's flooded with untested ones, which is what SkillBench and ToolBench are for. What happens to feedback loops when the agent calling your tool is doing another agent's job, not a human's, and nobody's left to ask how it went. The tradeoff between individual throughput and team-level understanding: my engineers ship more code than ever and know less about each other's work than ever, and code review alone doesn't fix that.