Project
ProductionProduction AI work
The applied-AI systems I built at The Wildcard Alliance: an LLM content pipeline on IBM Granite, a vision-model captioning service on ChromaDB, agents that playtest builds in CI, MCP servers in daily engineering work, and an adoption program that reached 80% of the team.
- IBM Granite
- LangChain
- ChromaDB
- MCP
- Sentry
At The Wildcard Alliance I built the studio's applied-AI systems and ran the program that got the team using them.
LLM content pipeline
A pipeline on IBM Granite models that streamlined asset creation for the studio's live-service title.
Vision-model captioning
A captioning service that describes images with a vision model and indexes the results in ChromaDB, so assets can be found by what they show.
Agents that playtest in CI
An autonomous agent framework, in Python with LangChain, that plays live game builds in CI and reports what it finds to Sentry. It gives the team automated playtesting coverage between human QA passes, filed in the same place as player-reported crashes.
MCP in daily engineering work
Model Context Protocol servers wired into the team's daily engineering workflows, so AI assistants work against real project systems rather than pasted context.
The adoption program
Tools only matter if people use them. I chose which workflows were worth automating, built the tooling, trained the team, and measured the effect on throughput. Adoption reached 80% of the engineering team.
Waypoint is where I took the lessons further, in the open: durable agent runs, human approval steps, and evaluation with a calibrated judge.