D BKN / 4.9TRHS / 8.3NODE / 29.7
Available · selectively · Q3 2026

Product,end-to-end.

I run AI-native B2B SaaS with you — discovery, spec, code, ship. Five years across product, the last two fully AI-native.

// Ask me anything
try:
// Analysis
SessionQUERY-01
Latency0.42s
Stress lvl12.4%
SoftwareSTABLE
Probability—
ThreadsACTIVE / 3
// Stress monitor
SubjectStable
§ 01

Currently building.

Live products · 2025–26
CHAPTER 001
● Live

Atingi — OKRs for franchise networks.

Multi-tenant SaaS for goal-tracking across hundreds of franchise units. Built 0→live as PO + technical lead in eleven months.
NEXT.JS · NESTJS · POSTGRES · BULLMQ · PGVECTOR · VERCEL AI SDK
↗
CHAPTER 002
● Live

Taillr — the AI Career OS.

CV tailoring loop: base CV in, role-adapted CV out, stitched by an agent ensemble. Structured billing, credit ledger, LGPD-native.
NEXT 15 · SUPABASE · DRIZZLE · AI SDK · INNGEST · STRIPE BR
↗
CHAPTER 003
● Open Source

Open Business OS — a starter for AI-native shops.

Template repo for multi-agent business operations. Slash commands, MCP servers, quality gates, parallel git-worktree delivery.
TYPESCRIPT · MCP · CLI · GIT WORKTREES
↗
CHAPTER 004
● Internal

Sollertis Agent Scripts — the orchestrator we use daily.

Cross-repo agent coordination layer. Custom slash commands, security and quality gates, parallel delivery via git worktrees and tmux.
CLAUDE CODE · TMUX · BASH · TYPESCRIPT
↗
§ 02

Atingi, by the numbers.

0 → live · 11 months
0
Modules
domain-isolated
0+
Endpoints
all OpenAPI-spec'd
0+
Migrations
zero-downtime
0+
AI agents
task + supervisor
0
Pages
App Router, RSC
0+
Components
design-system tokens
0
Specs shipped
spec-first delivery
0
Pricing plans
32 gated features
Built solo as PO + technical lead, 0 → live in 11 months.
§ 03

Selected work.

Three deep cuts

Atingi

OKRs that survive contact with a franchise network.

Goal-tracking SaaS for franchise networks running anywhere from twelve to eight-hundred units. The hard part isn't the OKR tree — it's the politics of cascading a metric down from holding to unit-manager and keeping the audit trail clean.

I ran it solo as PO and technical lead from a blank repo to production in eleven months. The tech is unremarkable on paper — Next, Nest, Postgres, BullMQ — and ruthlessly chosen on purpose. The interesting work was the agent layer underneath: a hundred-plus background agents that watch metrics, draft check-ins, and surface anomalies before the franchisee opens the dashboard.

What I learned: the spec is the product. Every feature here started as a written argument before a single ticket existed.

Atingi · Mapa estratégico · 2026
Visit Atingi ↗

Taillr

A CV tailoring loop, not another resume builder.

Most resume tools generate one document and pretend that's the job. Taillr models the actual loop: a base CV (the truth), a role description (the constraint), and a tailored output (the artifact) — re-runnable, versioned, billable per credit.

The agent ensemble is the product: extract, match, rewrite, lint. Each step is a separate model call with its own evals, and the user sees every diff. Brazilian payments are first-class — Stripe BR + PIX, credit ledger, LGPD posture from day one.

What I learned: structured billing is a moat. Plan versions, credit refunds, referral grants — it's all infrastructure most products bolt on too late.

Taillr · CV tailoring loop · 2026
Visit Taillr ↗

Open Business OS

The plumbing every AI-native shop is going to write twice.

An opinionated starter for teams who run their business the way I run mine: multi-agent, MCP-native, spec-first. It is the part of every project I rebuild from scratch — and shouldn't.

Includes the Sollertis agent script set as a reference orchestrator: slash commands, security gates, parallel delivery via git worktrees and tmux panes. Plus a small SDR Machine demo agent that scores opportunities and renders a prospecting radar.

What I learned: tooling is a moat too. Every hour saved compounding into the next ship is an unfair advantage by month three.

SDR Machine · prospecção AI-native · 2025
Visit the repo ↗
§ 04

Operating system.

How I actually work
PMs who can read the diff ship better products than PMs who only write tickets.

Daily tools

  • Claude Code as the primary surface
  • Cursor for surgical edits
  • MCP servers wired to Postgres, GitHub, Linear
  • Multi-agent on git worktrees
  • Vercel AI SDK for every product

Patterns

  • Spec-first — written argument before tickets
  • ReAct + RAG, not autonomous agents
  • Slash commands over chat for repeat work
  • Quality gates on every PR (lint, type, evals)
  • Flow-based delivery, no estimates

Beliefs

  • Velocity is a moat
  • Treat AI like infra, not magic
  • The boring parts matter most
  • Good PMs write code
  • The spec is a product
§ 05

How we work together.

Audit · Embed · Retainer

Three shapes of engagement. Same operator, three commitments.

1 day

Audit

Senior outside read on what to keep, cut, and build. For teams already shipping.

  • Live walkthrough of repo + spec
  • Architecture diagram, hand-drawn
  • 1-page tactical memo
  • 30-minute debrief call
Best forTeams shipping fast that want a sanity check
Request quote ↗
2 weeks

Embed

I sit in your stack, write the spec with you, deploy a real slice end to end. The handoff is the deliverable.

  • Spec written together, day one
  • Working code in production by day five
  • Agent infra that survives my exit
  • Handoff doc your engineers will not throw out
Best for0 → live on one of the right things, not all the things
Request quote ↗
Monthly

Retainer

Senior product + AI eng on call. ~6 hours synchronous a week, async PR review and spec writing in between.

  • Weekly 1:1 with founders
  • Async spec + PR review on critical paths
  • AI infra advisory
  • Hiring screen + onboarding for new product hires
Best forFounder-led teams that need a senior head without a full hire
Request quote ↗

Inside the two-week embed

  1. Day 1Sales calls + repo + spec read end to end
  2. Day 2One-page argument: ship this, not that
  3. Day 3Joint cut. Half the scope leaves the table.
  4. Day 4Smallest end-to-end slice scaffolded + deployed
  5. Day 5Customer-breakable build. Shared vocabulary.
  6. Week 2Agent infra hardened, handoff doc, team trained
§ 06

Background, condensed.

Five years, the relevant part
2025 +
Building products · Grupo Sollertis
AI-native B2B SaaS portfolio · Atingi, Taillr
Product engineer
2023 +
Cerq Company — AI consulting + product MVPs
Indie shop, founder operations
Founder
2022 +
DevSkin — enterprise product delivery
Consulting on product + agile shipping
Consultant
2021 – 22
Polen.me — led influencers + shipping squad
Globo Ventures-backed · +20% retention
Sr. PM
2020 – 21
TamoVip — promoted from BI to PM
First product role, learned the trade
PM

Education — PM3 AI Product Specialist + Product Master. Rocketseat full-stack JS. UNIFG IT Management.

Languages — Portuguese (native), English (working), Spanish (working). Agent reads all three.

Availability — Selectively, Q3 2026. Embeds two-to-eight weeks. Or full-time for the right thing.

NDA
Mutual · 2-page
IP
Transfers on payment
LGPD / GDPR
Compliant by default
Insurance
Liability covered
Bus factor
Documented handoff
// Mission brief

Embed with me for two weeks. Ship a real thing.

  • One spec, written together, ruthless about scope
  • Working code on day three, not week three
  • Agent infra that survives my exit
  • A handoff doc your engineers will not throw out
See the offer ↗
§ 07

Common questions.

Read this before emailing
Audit: same week. Embed: typically 2–3 weeks lead time. Retainer: depends on slot availability. Message and I'll come back within 24h with a real answer.
Yes, solo by design. Mitigated by: written-spec-first for everything we ship, paired sessions with one of your engineers from day one, and a handoff doc on day fourteen your team can drive without me. The handoff IS the deliverable. If I disappear tomorrow, you keep moving.
You do. Full IP transfer on payment. I keep no shadow code, no agent prompts, no specs. Mutual NDA is standard. Templates of process (not your work) stay with me.
Quote-based per scope. Send a message describing the engagement and I'll come back within 24h with a fixed price. No surprise invoices, no time-and-materials roulette.
Português (native), English (working), Español (working). Written deliverables in your team's primary language. Async + sync both available.
Claude Code as the primary surface, Cursor for surgical edits, MCP servers wired to Postgres / GitHub / Linear, my Sollertis agent scripts. I work in your repo, your CI, your stack. Nothing custom you have to maintain after I leave.
For embeds 4+ weeks at YC-stage companies, partly. Hybrid cash + equity considered case by case. Default is cash. I will not work pure equity for someone I have not met.
Two options after embed: (a) 30-day async tail at a flat rate for follow-up questions, or (b) drop into retainer. Both optional, no lock-in. If your team can drive it without me by day fourteen, neither is needed.
§ 08

Let's talk.

Open inbox