Private AI infrastructure for sensitive workloads.
- Role
- Co-founder
- Years
- 2026 — now
● Overview
A consultancy I co-founded to move sensitive AI workloads off public APIs: private, self-hosted infrastructure, security hardening and a documented handoff. We also build product for clients and ourselves.
● The hard part
Teams want LLMs on data they can't send to a third party. The hard part isn't the model, it's the infrastructure, the security posture and leaving the client with something they can run.
- 9
- adaptive site templates
- 1-click
- Railway deploys
- 2026
- founded
● What I built
- 01
Local Site Studio
Turns a Google business listing into a conversion-focused Next.js site. Places enrichment feeds an LLM that returns a validated SiteSpec, and a deterministic scaffolder builds the app across nine template families. Models never write executable code. One click deploys to Railway.
- 02
Veterans Education Benefits Map
An interactive resource explorer for The Service Connection Project, built on Next.js, NestJS and PostgreSQL with ID.me verification and hardened session handling.
- 03
Internal tools
A CRM on Next.js, NestJS and Prisma, a leads dashboard, and the VoiceMirror platform.
● Stack
- Next.js
- NestJS
- TypeScript
- PostgreSQL
- Prisma
- OpenAI / Anthropic
- Google Places
- Railway
- Docker