Shop AI platform
A licensed shop platform with an AI layer for service advisors and a diagnostics agent for technicians in the bay.
OutcomeAdvisor briefings generated from shop data with no model in the decision path.
- Python
- SQL
- Agents
- Evals
Applied AI · Parent of BLZ Foundry
DriveAIWerks is the parent company behind BLZ Foundry. We build governed AI systems for real operations: shops, product teams, and the people who run them.
01 · The companies
DriveAIWerks sets direction, standards and governance. Each company under it owns one part of the path from idea to running system.
Break the logic. Build the proof.
Takes an idea from rough problem to working proof: discovery, product definition, and the first build that holds up in front of users.
blzfoundry.com ↗Agents, evals, and the plumbing between them.
Where agent workflows, eval suites, and AI-assisted delivery get built and stress-tested before they reach a product.
See the work →02 · Selected work
A licensed shop platform with an AI layer for service advisors and a diagnostics agent for technicians in the bay.
OutcomeAdvisor briefings generated from shop data with no model in the decision path.
Took an automotive aftermarket kit from product to market: storefront, payments, order alerts, and a live ops dashboard for every order from paid to delivered.
OutcomeRetail, wholesale and international orders in one private dashboard, with every change on record.
Full brand and site for a classic and restomod shop: vector logo, chrome truck hero, challenge-coin build registry, and a self-serve edit page for the owner.
OutcomeThe shop updates builds, gallery and for-sale cars without touching code; every publish lands as a reviewed change.
03 · Approach
Four rules every system we build follows, from the first prototype to production.
The decision path runs on rules and data you can audit. Models sit on top, where they add judgment, not where they add risk.
Every agent ships with frozen test cases. If the eval goes red, the release stops.
Missing data, missing auth, missing context: the system stops and says so instead of guessing.
One repo, one main branch, one place decisions are logged. Every change has a record and an undo.