Technical Stack
Grouped by domain — and why each tool earns a place in my work.
Where models meet tools — retrieval, embeddings, and agents that stay on a leash.
APIs and services that other systems can depend on without drama.
Schemas, caches, and event stores that stay honest as volume grows.
Servers, containers, and delivery paths I am willing to wake up for.
Tracking and attribution that media teams can actually reconcile.
Interfaces that explain systems — and landing pages that do not hide the ball.
Engineering Principles
The rules I try not to negotiate away when building automation, ads pipelines, or infrastructure.
If it breaks at night, the design was incomplete.
I plan for retries, timeouts, and restore paths before I call something done.Automate the boring parts — keep people in the loop for the risky ones.
Agents and scripts get limited tools, clear permissions, and logs you can audit later.If we cannot measure it, we are guessing.
I prefer structured logs, clean event IDs, and health checks over dashboards that look busy.Security is easier when it is built into the first version.
Least privilege, secret handling, and input validation from day one — not as a cleanup sprint.Someone else (or future you) should understand it without a detective novel.
Clear boundaries, simple contracts, and notes on the decisions that matter.Clean data first. Clever optimization second.
Especially in ads and AI — bad events and weak schemas waste every optimization pass.
