Erik Mejia
I help teams move AI work from experiments to production faster, without trading away product judgment, reliability, or maintainability. I do the product thinking and the hands-on work: map workflows, design interfaces, write and review code, define evals, and ship AI systems people can actually use.
Proof
I build my own AI products from zero to one
Flowcost and Didastik are products I own and built from the ground up. They are where I apply the same judgment I bring to client work: turning messy AI capability into usable workflows, clear architecture, and product behavior people can trust.
Retainer
Q2 availability
AI Product & Design Engineer
Monthly
$10,000
Minimum
3-month engagement
What this includes
Workflow architecture
Map the product flow, model choices, retrieval, tools, handoffs, infrastructure, cost, latency, and failure paths before engineering time is committed.
RAG & knowledge systems
Turn company docs, data, and domain knowledge into retrieval workflows with source selection, grounding, and quality checks.
AI product UX
Design AI-heavy product behavior so users understand system state, trust outputs, recover from failure, and stay in control.
Production feature builds
Ship copilots, assistants, internal tools, agent workflows, and AI-powered product surfaces while modernizing brittle AI code paths.
Quality, cost & reliability
Define evals and success criteria, inspect failures, tune routing, caching, retrieval, and usage patterns, and reduce the risks that make AI systems hard to trust.
AI team tooling
Help your team use modern AI coding tools in real work: Codex, Claude Code, Cursor, Flowcost, and the habits needed to use them well.
Testimonials
People I have worked with
"Erik helped us move from a loose AI product idea to a clear, shippable direction. He brought rare range across product strategy, UX judgment, and engineering follow-through."
"Erik is strong at turning ambiguity into product decisions. He helped translate complex workflows into interfaces and systems the team could actually build, test, and improve."
Engagement
How the engagement works
The retainer is structured around delivery, not advisory theater. I work in your existing tools, make decisions concrete, and turn the important parts into shipped product.
01
Start inside the real workflow
I get into the product, codebase, data, team chat, GitHub, docs, and Linear or Jira so the work is grounded in what users and engineers are actually dealing with.
02
Pick the highest-leverage work
Each week, you bring product gaps, messy AI prototypes, unclear workflows, or technical blockers. I break them down, design the path forward, and ship the parts that need hands-on execution.
03
Leave the system stronger
Each cycle should leave behind shipped product, clearer architecture, better evals, stronger AI tooling habits, or a sharper decision about what not to build.
Book
Book a call
If the work sounds aligned, pick a time below. I use the first call to understand what you are building, where the AI work is stuck, and whether I can help directly.