AI Platform / Internal Tooling
ACB Platform
A sanitized case study of a multi-service internal platform that turns natural-language infrastructure requests into plans, Terraform artifacts, and operator-visible delivery workflows.
This public summary is intentionally sanitized. It excludes secrets, internal contracts, prompt details, RAG schemas, and protected operational workflows.
Challenge
The platform needed to connect prompt intake, planning, Terraform generation, observability, and feedback inside one operator-friendly workflow without exposing sensitive implementation internals.
Architecture snapshot
- Dashboard surface for prompt intake, plan review, and artifact visibility
- API layer coordinating requests, persistence, and workflow state
- Planning service converting intent into infrastructure plan structures
- Executor service rendering Terraform artifacts and local run paths
- Observability loop capturing health and issue signals across services
- Postgres plus vector-backed retrieval supporting history and feedback
Implementation approach
- Defined a platform flow that moved from prompt intake to Terraform output through a single orchestrated control path.
- Used containerized local workflows so the dashboard, API, planner, executor, and observability services could be started together for delivery and testing.
- Separated public-safe architecture framing from internal schemas and retrieval logic so the product could be discussed externally without leaking protected details.
Tech stack
- FastAPI
- Next.js dashboard
- Docker Compose
- Terraform
- Postgres
- pgvector
- Observability workflows
Want the implementation detail behind the public summary?
I can walk through the architecture choices, rollout tradeoffs, and the delivery outcomes that matter most for your team.