Available for new engagements
Most AI projects die in the gap between a demo that impresses and a system a business can depend on. Netraa is an independent engineering practice built to cross that gap — LLM applications, agents, and the unglamorous software that makes them reliable.
Services
Engagements are scoped to a decision or a deliverable, never to a headcount slot. If the honest answer is that you do not need any of this, that is a fine outcome for a first call.
From a prototype that impresses to a system you can put a customer on.
Chat and copilot interfaces, structured extraction, document pipelines, evaluation harnesses. The model is the easy part — the work is everything around it: prompt versioning, fallbacks, cost ceilings, latency budgets, and knowing what to do when the model is confidently wrong.
Getting the right context in front of the model, every time.
Ingestion, chunking strategy, hybrid search, reranking, and the boring discipline of measuring retrieval quality instead of guessing at it. Built on whatever your data already lives in — Postgres and pgvector, Azure AI Search, or a purpose-built index.
The application, the API, and the pipeline that ships it.
Full-stack product work in TypeScript and .NET. Greenfield builds, legacy systems that need a second act, and the integration layer between the two. Typed end to end, tested where it matters, deployed on infrastructure your team can actually operate.
A senior engineer in the room when the decision is expensive to reverse.
Architecture reviews, build-versus-buy calls, AI readiness assessments, and technical due diligence. Short engagements with a written deliverable — a recommendation you can hand to a board, not a slide deck full of hedges.
Approach
The failure mode in AI work is not writing bad code. It is spending six months building the wrong thing beautifully. Every engagement is structured so you find out early.
A week of discovery before a line of code. What is the actual failure mode you are trying to remove, what does success measure out to, and is AI even the right tool here? Sometimes the answer is a SQL query and a cron job, and I will tell you that.
One narrow path, end to end, in production, behind a flag. Real data, real users, real cost per request. A thin slice that works beats a broad plan that has never been tested against reality.
Observability, evals, runbooks, and documentation written for the engineer who inherits this after I am gone. The goal is a system your team owns — not a dependency on me.
Engagements
A focused review of a system, a codebase, or an AI initiative, ending in a written recommendation with costed options.
Scoped delivery of a working system, shipped in slices, with your team involved from the first week so the handoff is not a cliff.
Part-time senior capacity on your team — architecture, code review, and hands-on delivery at a defined weekly commitment.
Stack
Nothing here is novel, and that is the point. Your team has to run this after the engagement ends.
About
Netraa is a small, deliberately independent practice run by Andrew Gunn, a senior full-stack engineer who spends his days as the technical owner of a corporate web platform — enterprise .NET, React, SQL Server, Azure, and the CI/CD that holds it together.
That background matters more than it might sound. Most AI advice on the market comes from people who have never had to keep a production system up on a Monday morning. The interesting problems in this field are not model problems — they are integration problems, data problems, and organizational problems wearing a model-shaped costume.
You work with me directly. No account manager, no bench, no handoff to someone more junior after the sales call. That caps how much work I can take on, which is a feature: I only take engagements I can genuinely do well.
A short email is enough to start. If it is not a fit, I will say so quickly and point you somewhere better.
Typical reply within one business day