Consulting · AI

AI engineering, from spec to eval.

I help companies and tech teams put AI to work safely — end to end, or wherever it adds the most value — with Spec-Driven Development (Spec Kit), agent harnesses, MCP and evals to measure quality.

At EBANX — one of the pioneers of AI engineering with Devin — I've spent the past 8 months working in a cutting-edge AI-first model on payment systems for global merchants. Along the way I received training from Cognition and Claude Code mentoring from Objective's specialists, and took the international Claude Certified Architect – Foundations (CCA-F) certification. I also completed the Tech Leads Club course with Waldemar Neto on AI Architecture, Harness and Agentic Factories.

Talking with colleagues, former colleagues and friends, I noticed that many companies — and many tech professionals — still aren't capturing the performance and quality gains of agentic factories. That practice, proven in production, is what I bring to your team.

It starts with an initial diagnosis (R$ 10,000 BRL) and a roadmap of the AI best practices I've consolidated over the last year.

Send an e-mail
Raphael Bohrer wearing sunglasses in a car selfie

01 — Where I help

Where I help

Spec-Driven Development

Specs as the source of truth: from requirement to plan and tasks, with Spec Kit, so agents work with clear context and boundaries.

  • SDD
  • Spec Kit

Agent harness & MCP

A working environment for agents such as Claude Code and Devin: tools, permissions and MCP integrations with your team’s systems.

  • Claude Code
  • Devin AI
  • MCP

Evals & Loop Engineering

Evals to measure the quality of what agents deliver, and feedback loops to keep improving the process.

  • Evals
  • Loop Engineering

Agentic factories

Orchestrating agents in repeatable workflows, with solutions architecture that keeps quality and safety in production.

  • Agent orchestration
  • Solutions architecture

02 — How it works

How it works

  1. Discovery

    Understand the team's design system, processes, responsibilities and workflow, identifying bottlenecks and opportunities for improvement.

  2. Analysis

    Analyze specs, harness, skills (skill augmentation), playbooks, evals, cohesion and coupling, ambiguity, sensors, context and memory gaps, loops and validations — or the lack of them.

  3. Roadmap

    Deliver, as a Pareto on a scale of value, a diagnosis with recommendations and direction to make your company more competitive and keep it from falling into obsolescence.

Shall we talk about your team?

Share the context — team size, stack and goal — and I'll reply personally.

R$ 10,000 (BRL)Initial diagnosis and roadmap of AI best practices

Résumé