# David George Smith

> AI Orchestrator & Technical Product Lead. I build the AI-powered systems engineering teams run on — MCP integrations, LLM automation pipelines, Copilot adoption dashboards — and lead the people who use them. I sit at the layer above the automation: I set strategy, manage engineers, own product decisions, and measure what AI is actually worth.

Roles: Senior Full Stack Engineer, Engineering Manager, Product Owner, AI Enablement Lead. Current employer: Solidigm. Education: B.S. Computer Science, Sonoma State University.
Contact: 1426davejobs@gmail.com · https://linkedin.com/in/codingforgood · https://github.com/smithdavedesign
MCP server (ask about this person from any MCP client): https://smithdavedesign.herokuapp.com/v4/mcp

## Measured impact
- 25% — License optimization identified. Copilot seat usage analysis surfaced a 25% license optimization opportunity.
- 15–35% — Operational cost reduction. Documented 15–35% operational cost reductions across engineering teams using AI tooling.
- 10+ — Internal agent tools shipped. Custom internal agent tools deployed by engineers trained through the AI KT Series.
- 40% — Less developer rote work. Reduction in repetitive developer work from MCP-based agent tooling.

## Case studies
- [The Governance & ROI Engine](https://smithdavedesign.herokuapp.com/v4/work/copilot-roi-engine/): Engineered a real-time dashboard for Copilot usage reporting and analysis — benchmarks, adoption trends, team-level comparisons and an ROI calculator built for CTOs and engineering managers. Tooling: Telemetry ingestion, Custom visualization layer, TypeScript, React, ROI model. Result: Identified a 25% license optimization opportunity and documented 15–35% operational cost reductions across engineering teams.
- [Building the Digital Workforce](https://smithdavedesign.herokuapp.com/v4/work/digital-workforce-kt-series/): Architected an org-wide Knowledge Transfer series training engineers to build Model Context Protocol servers, Skills, and custom instructions — the human infrastructure that makes AI adoption stick. Tooling: Model Context Protocol, Claude Code, GitHub Copilot, Custom Instructions, Skills. Result: Accelerated the deployment of 10+ custom internal agent tools, reducing developer rote work by 40%.
- [AI-Powered Developer Tooling](https://smithdavedesign.herokuapp.com/v4/work/ai-sdlc-tooling/): Developed a suite of AI-integrated coding tools that streamline repository management and pull-request review. Tooling: LLM agents, Map-Reduce-Align, GitHub API, Node.js, MCP. Result: Shorter review cycles and consistent PR quality through structured, evidence-backed agent recommendations.
- [Figma → Jira / Wiki Automation](https://smithdavedesign.herokuapp.com/v4/work/figma-to-jira/): LLM + Figma MCP + API integration that auto-generates Jira tickets and technical wiki documentation directly from design context. Owned end to end: product definition, stakeholder alignment with PMs and Scrum Masters, and engineering execution. Tooling: Figma MCP, LLM, Jira API, Confluence, Node.js. Result: Eliminates manual documentation labor between design and engineering across the org.
- [QA Test Case Automation](https://smithdavedesign.herokuapp.com/v4/work/qa-casegen/): Automated test case generation pipeline via Storybook and MCP servers, reducing manual QA effort while maintaining quality standards team-wide. Tooling: Storybook, MCP servers, LLM, QA tooling. Result: Reduced manual QA effort while keeping quality standards consistent across teams.

## Products built
- [RepoHQ](https://smithdavedesign.herokuapp.com/v4/builds/repohq/): RepoHQ syncs all of my GitHub repositories, scores their health, and turns that into ranked, quantified next actions. An AI advisor proposes the work; a sandboxed agent factory runs it on free models and judges every change before it lands. Stack: Next.js 16, TypeScript, Neon Postgres, Drizzle ORM, Auth.js (GitHub OAuth), BullMQ / Redis, Claude & Gemini, MCP, Vercel. Source: https://github.com/smithdavedesign/Github-HQ · Live: https://repohq.vercel.app
- [AI Radar](https://smithdavedesign.herokuapp.com/v4/builds/ai-radar/): AI Radar crawls six public sources every week and turns raw signals into a RadarScore (0–100) for each AI tool, with sub-score breakdowns, score history, movers and head-to-head comparisons. A Claude enrichment agent writes the comparisons. Stack: Next.js 16, TypeScript, Neon Postgres, Drizzle ORM, Inngest, Claude API, Recharts, Playwright, Vercel. Source: https://github.com/smithdavedesign/AI-Trend-Tracker · Live: https://ai-trend-tracker-psi.vercel.app
- [Open Adventure](https://smithdavedesign.herokuapp.com/v4/builds/open-adventure/): Open Adventure (working name Travel Roamer) is a discovery platform for outdoor trips. A full-screen map with vibe and trip filters leads to destination and trail pages built on a provenance-backed content model, where every fact is sourced, attributed and freshness-gated. Stack: Next.js 16, TypeScript, Postgres + PostGIS, Prisma, MapLibre / MapTiler, Auth.js (Google), Sentry, Playwright. Source: https://github.com/smithdavedesign/go-adventure · Live: https://go-adventure-three.vercel.app
- [Open Fly Travel Assistant](https://smithdavedesign.herokuapp.com/v4/builds/open-fly/): A collaborative travel app: plan trips together, drop in PDFs or screenshots of booking confirmations and let Claude extract the events, then split costs across the group with real-time sync. Stack: Next.js, TypeScript, Supabase, Claude API, Mapbox, Tailwind v4, Vercel. Source: https://github.com/smithdavedesign/Open-Travel · Live: https://open-travel-azure.vercel.app
- [Roots & Branches](https://smithdavedesign.herokuapp.com/v4/builds/roots-and-branches/): A full-stack genealogy platform: build and visualize a family lineage as an interactive graph, map ancestors' journeys, attach photos from Google, and collaborate with relatives — with subscriptions, AI biographies and role-based access built in. Stack: React + Vite, Node.js / Express, Supabase Postgres, React Flow, Stripe, Google OAuth, Vercel / Render. Source: https://github.com/smithdavedesign/family-tree · Live: https://familytree-e.com

## Governance approach
Autonomy is earned, scoped and reversible. Every agent I deploy sits at an explicit level on this ladder, with a named human above it.
- L0 Suggest: Agent drafts; human writes. Default for anything customer-facing. (human checkpoint: Every output)
- L1 Draft & route: Agent produces an artifact and routes it to a named reviewer (tickets, test cases, PR reports). (human checkpoint: Before publish)
- L2 Act in sandbox: Agent executes inside a scoped environment with gated skills and logged tool calls. (human checkpoint: Before merge / deploy)
- L3 Act with hard stops: Agent runs end to end; defined hard stops (security, money, people data) always escalate. (human checkpoint: On hard stop)

## Stack
Model Context Protocol, Claude / Claude Code, GitHub Copilot, LangGraph, Next.js, Tailwind CSS v4, TypeScript, Node.js / NestJS, Python, AWS / Docker, GitHub Actions, Snowflake / Redis / Mongo

## History
- Solidigm: Engineering Manager, Product Owner and AI Enablement Lead — AI tooling, MCP integrations, Copilot ROI analytics, org-wide AI curriculum.
- Madison Logic: Modernized large marketing engineering platforms, dispatchers and front-end systems.
- UC Davis: Built internal tools and systemized documentation.

Resume: https://smithdavedesign.herokuapp.com/v4/resume