Idir Z.

I build backends and automate things in Python, and my background is in testing systems rigorously. I spent six years at Ciena as a verification engineer finding failure modes in mixed-signal silicon, writing the Python and Jenkins automation an entire engineering team relied on daily. More recently I've applied the same approach to AI systems, where the question shifts from "does the circuit behave correctly" to "does the model, and what happens when it doesn't." Most of my recent work is on the reliability side of LLM applications. It's easy to wire up an API call. It's harder to make sure the feature still behaves when the model is down, returns nonsense, or quietly burns through your budget. On CoachIQ, an AI coaching platform I co-built, I designed a three-tier fallback across Anthropic and OpenAI with a static backstop, cost caps enforced at the database level, and an architecture where the model only writes user-facing prose while the underlying logic stays deterministic. A model failure cannot produce a wrong result for a user. I built the rest of that product too: an Express and TypeScript REST API on Postgres serving a Next.js web app and a React Native mobile app, with row-level security across every user table and CI/CD on GitHub Actions. What I can help with: Python automation, testing frameworks and data pipelines Evaluating and benchmarking AI output, including validation, failure analysis and regression testing LLM integration that fails safely, including fallback logic and cost controls Backend services and REST APIs in Node, TypeScript, Express and Postgres CI/CD pipelines in GitHub Actions, Jenkins and Docker Happy to discuss what you're working on and whether I'm a good fit.