Igor Paes Ferreira
Paes Ferreira
Open to AI roles · Remote · Hybrid

I build AI products
that work in production.

From discovery to deploy: I combine product vision, analytical depth and business sense to turn complex problems into software that delivers value. Electrical engineer, 7 years at Volvo, 8 as a founder in the energy sector.

  • eXmeshMAS · agentic intelligence layer
  • eTradeflowETRM · agile energy contracts
  • eFlowingCRM · faster free-market migrations
LangGraphClaude · Vertex AINext.jsKubernetesGCPPython
The Journey

From Bosch China to AI in production

19 years of building. 3 countries. One consistent thread.

Click any card to expand

2007

Living and working in China at 21 — immersed in a completely different culture while engineering production automation systems alongside Chinese teams. The scale of everything was humbling. The seed of thinking beyond borders was planted.

2007

Stuttgart, the birthplace of the automobile. Working inside Bosch's R&D labs taught me what it means to engineer for reliability — no shortcuts, no approximations. German precision became a personal standard I still hold.

2010–2017

Seven years at Volvo shaped how I think about systems, reliability, and accountability. Starting as an engineer and ending as a product leader — I learned that great products are built by people who understand both the machine and the human using it.

2011

An electrical engineering degree is fundamentally about modeling systems — inputs, outputs, feedback loops, failure modes. That mental model still drives how I architect software, teams, and companies.

2017

December 2017. Stanford, Google, YC — meeting founders who were actually shipping world-changing products. I flew back to Brazil with one clear thought: I'm going to build something. That trip didn't inspire me — it decided me.

2018

The first office was a single room. Just a desk, a laptop, and a lot of uncertainty. The FOHAT logo on that black wall was the only thing that felt real — a holding under construction.

2018–2023

Five years of building complex B2B products in regulated industries. Every feature we shipped had to survive ANEEL regulation, CCEE compliance, and the scrutiny of traders who measure money in megawatts. Today eXmesh agents monitor the migration pipeline end-to-end, detecting bottlenecks before they become delays — without taking humans out of the decisions that matter.

2022–2024

I got certified as an Innovation Board Member by Gonew.co — 100 hours, 40 of them in Board Real Practice — and served three consecutive terms on OSINOVA's Board of Advisors. I brought both lenses to the table: the traditional corporate world and the startup builder's.

2024–2026

MUTHUR receives any demand and distributes work with no human intermediary. Bishop holds the organizational memory — it reads our documentation and historical decisions and detects drift before it becomes a problem. Walter runs engineering end-to-end and was the first to autonomously merge code into production. Each area has its own specialized sub-swarm, coordinated through the central mesh — and every action passes a Governance Gateway that validates OKRs, mandates and decisions, human in the loop. What we built internally is being shaped into a product.

2026

The companies I founded are in GTM and being prepared to become AI First. I'm now open to channeling everything I've built — the AI systems, the product sense, the execution discipline — into a role where I can make an immediate, measurable impact. Remote. Hybrid. Brazil and international.

AI Systems · Production

eXmesh — Multi-Agent System, a mesh of coordinated agents running in production on GCP.

MUTHUR orchestrates. Bishop strategizes. Walter executes. 4 Synthetics monitor. Human-in-the-loop preserved.

MUTHUR

Mother OS · Orchestration Layer

Activated by chat. Routes to specialized manager agents, extracts memory from conversations and keeps everyone aligned to OKR-based governance.

ClaudeCloud RunGCSVertex Search

Bishop

Strategy-Ops AI

Reads ADRs, BRS, and SPECs, detects architectural drift, analyzes strategy documents. Bilateral conversation via chat HITL.

LangGraphClaude · Vertex AIPostgreSQL HITLLangfuse
🏆 1st autonomous PR merge in production

Walter

Surgical Executor

Our first agent to autonomously merge a PR in production. Receives task openers, runs tests, opens PRs and self-declares ready for merge.

LangGraphDual-modelBitbucket APICloud Run Jobs

Synthetic Workers ×4

Infra Monitors per Squad

Four autonomous agents — one per product squad. Each runs headless on Cloud Run, commits audit trail, and posts status to the FIPA Bus.

Claude Agent SDKCloud RunGCSGKE

Architecture · eXmesh Platform

Real control flow — from human to GCP substrate. Governance Gateway ensures every dispatch respects OKRs and Mandates before reaching agents.

HUMANOScrons · OKRs (origem)MUTHURmain-swarm (mother) · orquestradorADR-022GOVERNANCE GATEWAYOKRs · Mandates · Decisions — todo dispatch atravessa aquiADR-025 · ADR-026TIER 1 · MANAGERS — DECIDEMADR-003 · ADR-022HelenaSales OSSonnet · hot-pathEduardoMarketplace OSSonnet · hot-pathKatarinaGrowth OSSonnet · hot-pathWalterOperations OSSonnet · hot-pathBishopStrategy OSOpus 4.7 · long-horizonfan-out · Send API (in-process) / Redis Streams (cross-deployment)ADR-002 · ADR-012TIER 2 · WORKERS (SYNTHETICS) — HEADLESS, SEM AUTORIDADEADR-012 · ADR-020bxsalesgrafo fixo · Haikubxcrmgrafo fixo · Haikubxmktgrafo fixo · Haikubxgwtgrafo fixo · Haiku+ beatriz (SDR email) · runner (scheduler)6 MCP SERVERS · INTERFACE DE FERRAMENTAS · in-clusterfleet:8001 · memory:8002 · jira:8003 · bitbucket:8004 · governance:8005 · fipa:8006merge_pr = controle na interface (R4): destination='test' + mandate constraint, fail-closedADR-026 · ADR-025GOOGLE CLOUD · GKE · ArgoCD (GitOps) · Memorystore / RedisGKE us-east4 · Cloud Run us-east5 · project microservices-fohat · ns beenx-prod/testADR-007 · ADR-024resultados → Decisions
controle / orquestraçãoferramentas / dadosgovernançainfra / substrato

← swipe to see the full diagram →

Platforms Built

Production software used by real companies.

Let's build something.

Available for AI Product Manager and Technical Product Manager roles. Remote/Hybrid. Brazil and international.