Forward Deployed Engineer · Full-Stack · AI Workflows · Customer Engineering
I embed with teams, ship AI-native workflows, and own the outcome end-to-end.
- 🤖 AI agents for enterprise CS teams at Cora — multi-step reasoning agents that connect to Salesforce, Slack, BigQuery, and email to automate workflows like EBR creation, account health scoring, and renewal management. Production system with 90% repeat usage.
- ⚙️ Cursor rules, skills, and hooks for convention-heavy codebases — a structured approach to onboarding engineers faster using AI-native development tooling. See the system →
- 🚀 Production AI workflows — full-stack systems that combine LLMs, evals, and real integrations (Salesforce, Stripe, email) into something teams actually use. At enterprise scale: 11.4 sessions/user, EBR prep faster and more detailed, at-risk accounts flagged automatically.
- 🧑💻 Languages & frameworks: TypeScript / Next.js / React / Python / FastAPI / Vite
- 🗄️ Data: PostgreSQL / Neon / Supabase / MotherDuck / DuckDB / SQLMesh / dlt / BigQuery / Snowflake / pgvector
- ☁️ Cloud & infra: AWS (Lambda, S3, SES) / GCP / Vercel / Render / Hetzner / Cloudflare / Docker
- 🧠 AI: LLMs in production (OpenAI, LangChain, Agno, evals, Langfuse, prompt engineering) / Cursor (rules, skills, hooks, agents, MCP)
- 🔌 Integrations & ops: Salesforce / Slack / Gong / Stripe / Inngest / Sentry / Redis / Upstash / Auth0
I start with the real problem, not the interesting one. Ship a working version in days, harden it over weeks, hand it off so it runs without me. Every system I build is designed to be owned by someone else after I leave.
| Repo | What it is |
|---|---|
| spec-to-pr-agent | Spec → PR agent — multi-step planning with validation gates, convention-aware codegen, programmatic diffs, evals, CLI + dashboard |
| repo-health-agent | Production-hardened GitHub repo health scoring — deterministic scores, optional LLM explanations, fixtures, evals, CLI + dashboard |
| diffusers | Cursor rules + skills + hooks for onboarding engineers onto a convention-heavy library (HuggingFace diffusers) |
| ai.rules.prompts | Reusable Cursor rules, hooks, and skills for common development workflows |