Fractional Founding Engineer
For founders without a CTO: build a first version, rescue a stalled project, or prepare a prototype for real users. Fixed scope, published pricing.
Prefer to start with a message? Tell me about your project
Free
US$ 3,000
from US$ 15,000
Start with the product or process you need to improve. I help define a useful first delivery and how we will check that it works.
Turn documents, spreadsheets and disconnected tools into a workflow with a history of changes and a review step for uncertain results.
Example scope: read a document, extract its data and check the resulting record.
See how to startYou hired an agency or built it yourself, but the product is stuck. I assess what can be reused and which fixes matter first.
The starting point: access to the code, environment and failing user journey.
Explore project rescueYour prototype works in a demo. Now check permissions, integrations and the main user journey before bringing in real customers.
A reusable codebase and a bounded scope may fit the Production-Ready package.
See what is includedA customer portal, B2B SaaS or internal tool for a specific need. We agree on the first user journey and what makes it ready to ship.
The Sprint delivers scope and a clickable prototype. Implementation is a separate stage.
Compare stages and pricing
Since 2021 I've worked across almost every role in a data and AI product: ML research, full stack, data science, ML engineering and product management. On an AI product for container terminal operations, I built the model as the ML engineer, then took on the PM role while still writing production code.
I'm a generalist by design. I go deep enough in each layer to ship it and keep it running, and I know when a problem needs a specialist. I've worked remotely, in English, at a consultancy serving European companies. I'm in São Paulo (UTC−3), which overlaps US and European hours.
Loopt's AI product replaces manual yard planning. According to Loopt, the result was −27% unproductive moves and +21% productivity.
Public data · MundoLogística, Jan 2026
Read the full case study →In my own account, as the fintech's first data hire, I built the credit analysis model: a process that took 48 hours on average dropped to under 2 hours.
My own account
Read the full case study →An app with an AI advisor that logs money, workouts and sleep from one sentence. Built by me, end to end.
Read the full case study →I wrote "How to Build ML Model Training Pipeline" for the neptune.ai blog (2023).
Prior professional experience and a personal project, not studio client engagements. Self-reported results have no external publication.
You know the price before you sign. Every step ends with something you own, and you decide whether to take the next one. The first call is free and may end with "don't build software yet."
Free
A call to understand the problem and whether it's worth building now.
Payment: No cost, no commitment.
US$ 3,000
A build-ready plan for your v1, with a fixed price attached.
Fully credited to your MVP
Payment: 100% upfront.
Fully credited toward your MVP if you sign within 30 days. The deliverables are yours either way.
US$ 7,000
For when you already have a prototype and need it in production.
Payment: 50% at kickoff, 50% on delivery.
No rewrite and no new features; those go into an MVP or Evolution.
from US$ 15,000
Your product in production, built to the scope from the Sprint.
Payment: 40% at signing, 30% at staging beta, 30% at go-live.
Custom quote
Everything in Core, plus more roles, more integrations and one big feature.
Payment: 40% at signing, 30% at staging beta, 30% at go-live.
from US$ 1,500/mo
After launch, the same engineer keeps owning your product.
Payment: Monthly, 100% upfront.
| Plan | Hours/mo | Price/mo |
|---|---|---|
| Essential Monitoring, fixes, security updates and one monthly call. | ~10 h | US$ 1,500 |
| Evolution Essential plus a shipped feature every two weeks. | ~25 h | US$ 4,000 |
| Technical partner Evolution plus roadmap, fundraising support and your first hire. | ~45 h | US$ 7,000 |
Billed monthly in advance, 3-month minimum. Unused hours don't roll over. Extra hours: US$ 160/h.
Software projects fail on scope that grows without a price, on dependence on whoever built it, and on payments that don't track delivery. These rules exist for that.
After go-live, I fix anything that doesn't meet the acceptance criteria at no cost. New features aren't bugs: they go to Evolution or a change order.
The Sprint defines what's in and what's out. Changes mid-project become a priced change order, or a one-for-one swap at no cost.
The repo, database and infrastructure live in your accounts from the start. IP transfers to you on final payment.
You pay per delivered stage. The next stage starts once the previous one is paid. No surprises on either side.
I never run two MVPs at once. When I take your project, it has my full attention.
Every week you see working software, not slides. Course corrections happen early.
I work AI-native: coding agents write and review alongside me, and tests and infrastructure are automated from week one. That frees my time for the decisions that need judgment.
I use AI agents to write, test and refactor. Every line that reaches production goes through my review and the test suite.
From the start: tests on the core flow, automated deploys, staging separate from production, monitoring with alerts.
Architecture decisions, runbooks and handoff docs live in the repo. Any engineer can pick up where I left off.
Pentesting, brand design, native mobile or legal: I bring in a specialist, quoted separately and agreed with you first.
You always talk to me. No account manager, no handoffs.
Better to say it now than halfway through the project.
Yes. We start by understanding one workflow, its cost today and how to check an improvement. If the solution still needs defining, the Sprint delivers scope and a clickable prototype, not a production automation. Implementation is quoted separately.
I can assess the existing code and environment with authorized access. A reusable foundation and a bounded set of fixes may fit Production-Ready. A rebuild or new features need a separate scope and proposal before work begins.
The Sprint takes 2 weeks and Production-Ready 2 to 3. An MVP Core goes live in 6 to 7 weeks, an MVP Plus in 8 to 10. The milestone schedule comes out of the Sprint, along with the fixed price.
You do. The repo, database and cloud live in your accounts from day one, and IP transfers to you on final payment. Sprint deliverables are yours even if you don't continue to the MVP.
Your code, repo and cloud live in your accounts from day one, and documentation is written throughout the project, not at the end, so any senior developer can take over. Schedules carry a 15–20% buffer. And if the project can't continue, you get a pro-rata refund for anything paid and not delivered.
An agency makes sense if you need a large team for the long haul. A full-time hire makes sense once the product has traction. Before that, you get one senior engineer who owns the build, at a fixed price set after the Sprint.
They get you a prototype, and that's often a great start. I deliver production: protected data, auth, integrations, deployment, monitoring, and someone accountable when it breaks. If you already have the prototype, that's Production-Ready.
Yes to both. I can sign the NDA before the first call. In your project, you are the controller of your product's data and appoint your own DPO where required; I act as your processor, under your instructions, and the DPA puts that in writing. Personal data stays in infrastructure under your accounts.
I invoice in USD from my Brazilian company, and you pay by international bank transfer. I work B2B only: EU companies need a VAT ID (reverse charge applies), and US companies receive a W-8BEN-E from me.
For 60 days I fix bugs at no cost. After that, you can carry on with the docs in your repo, or keep me on through Product Evolution, with capped monthly hours.
Tell me what you're building. I'll tell you honestly whether to build it now, which step fits and what it costs. If the answer is "not yet", you'll hear that too.