Software engineer shipping language-model-powered products.
Four years of writing the unglamorous code behind AI features that work in production — backends in FastAPI and Laravel, interfaces in Next.js, model orchestration that doesn't fall over at 3 a.m.
I build the systems that sit between people and large language models — APIs, agents, retrieval pipelines, and the interfaces that make all of it feel sane.
Most of my work lives at the seam between Python and TypeScript. I reach for FastAPI when latency and streaming matter, Laravel when a real product needs to ship without yak-shaving, and Next.js + Tailwind + shadcn/ui when the front end has to feel as deliberate as the backend.
I care about the parts most demos skip: retries, evals, cost ceilings, tracing, the moment a model returns garbage and a real user is waiting. Docker and Coolify keep my deployments boring on purpose.
A chat-first assistant that routes prompts across OpenAI, Anthropic and Gemini based on cost and task, with streaming responses, tool calls and a usage dashboard for teams. Built end-to-end.
Retrieval-augmented search over thousands of contracts. Hybrid BM25 + embeddings, citation-grounded answers, audit logs and a permissions model lawyers actually trust.
A modular e-commerce platform powering several mid-sized stores — payments, inventory, admin tooling, and an AI product-description generator wired into the editor.
Got a hard problem with LLMs, a backend that needs shipping, or a product idea you keep coming back to?
I'm taking on a small number of new engagements. Tell me what you're building — I'll reply with a real opinion, not a sales pitch.
$ open mailto:hello@ifaz.dev→