Building an AI Team from Zero

The situation

A supplement manufacturing company wanted to bring AI into their operations. They had ideas — a chatbot that knows their entire product catalog, an email tool that writes in their voice based on past correspondence, integrations between their internal systems. What they didn't have was anyone who could build it.

No AI team. No tech stack chosen. No architecture. No idea how to evaluate AI developer candidates. They needed someone to come in, set everything up, and leave them with a functioning team that could keep building without outside help.

What I did

Hiring

I screened and interviewed 20 candidates and hired 2 AI developers. The challenge wasn't finding people who could code — it was finding people who could work in a small team, handle ambiguity, and build production systems (not just notebooks). I designed the interview process, wrote the technical assessments, and made the final hiring decisions.

Architecture & tech stack

I defined the full technical architecture from scratch — what models to use, how to structure the RAG pipeline for the knowledge chatbot, how to handle the email tool's need for style matching, and how to design the system so future integrations could plug in without rewriting core components.

Product roadmap

I planned the first 6 months of development: which product to build first (the knowledge chatbot — fastest to ship, clearest ROI), what to build second (email writing tool), and what to defer (system integrations — dependent on internal API readiness).

Training & handoff

For 3 months I mentored the team daily — code reviews, architecture discussions, debugging sessions. I communicated directly with company stakeholders so the developers could focus on building. By the end, the team was operating independently: they knew the stack, understood the patterns, and had a clear roadmap to follow.

Why this engagement model works

Most companies that want AI capabilities face a chicken-and-egg problem: you need AI expertise to hire AI people, but you don't have AI expertise yet. Bringing in a senior engineer for 3-4 months to bootstrap the team solves this cleanly — you get the right hires, the right architecture, and the right habits from day one, without committing to a long-term consulting engagement.