I code agentically to gain clarity and depth on anything I tackle, then ship what works. Strategy is intent; code is reality—that's why I build to validate, not theorize. I use data when the path is unclear and trust judgment when it's time to act.
Director at Adobe Commerce, building a baseline suite of interactive prototypes that collectively represent the entire commerce platform and customer experiences. This enables teams to adopt prototype-first product development—exploring workflows through working examples, instead of parsing documentation. I saw teams internalize patterns faster when they can interact with prototypes. That's why I'm now helping teams code their way to certainty before committing at scale—keeping AI use principled and responsible in real-world systems.
Work repos are internal to Adobe, but here's what I can share: Helpful Agents captures some of the workflows and frameworks I use—adapted so they're infrastructure-agnostic and immediately useful. I value meaningful exchanges—let's trade notes on agentic workflows or what you're building with AI.
Outside of work, I'm focused on steering AI to shape the next generation. I want to help them become critical thinkers with grit, life skills, and the wisdom to use AI safely—not as a shortcut that weakens judgment or self-reliance. I build with my kids first (music, sports, arts, literature, science). It's a work in progress.
These projects share a common thread: they're built through agentic coding workflows, and those workflows themselves reveal patterns worth capturing:
Inspiration — If you treat Cursor as a thinking partner (not just autocomplete), you've been having months of conversations about what matters to you. Inspiration turns that history into a mirror: Theme Explorer reveals patterns you didn't see ("I've been circling this problem in 12 different conversations"), and Knowledge Graph extracts entities and relationships. Reads Cursor's local chat database (reverse-engineered), with optional pgvector sync for massive histories (2GB+). What surprised me: the recurring themes that emerge across months—ideas worth building, insights worth sharing.
Next up: Exploring how to build a useful knowledge graph from my own AI chat—what entities and relationships actually matter?
Allegro — A music practice companion to cultivate the love for music. I learned that audio recognition trained on studio recordings fails completely when kids stop + repeat mid-phrase and play wrong notes. Current challenge: most APIs work on full songs (multi-instrument + vocals), but I need to identify songs from single-instrument playing (just piano or guitar). Open to ideas!
Next up: Real-time "play along" suggestions.
LifeSynced — Unified calendar with a privacy-first design. The "[Personal Event]" placeholder with password reveal made it safe to demo to any audience.
Dad Aura — Real-time aura scoring from my son's Apple Watch. What surprised me: him controlling score "flips" created more engagement than the scoring itself.
Next up: AI behavior that knows when to say no and nurture healthy values.
Cursor + Claude (Opus for coding, Sonnet for writing) — I focus on product requirements and architectural decisions; Cursor handles syntax and implementation patterns. This workflow lets me build at the speed of thought, iterate based on actual use.
For implementation, I use Next.js, TypeScript, Python, PostgreSQL, Vercel—guided by Opus based on latest best practices. For agent architecture: LangChain + MCP. Framework choices matter less than shipping validated solutions quickly.