PITANGA LABS

Applied AI lab · Porto Alegre → 2026

The kernel companies run on.

We build the kernel that writes into your systems of record, and stay until it runs the work.

The pitanga · drag to turn · click to squeeze

Most of the workday isn’t the work.

58%

of a knowledge worker’s day goes to coordination: chasing status, reviewing documents by hand, and keying the same record into two or three systems.

Hiring buys more of the 58%. Headcount scales; margin does not.
Asana, Anatomy of Work Global Index 2023 · 9,615 workers

Every company can now buy intelligence at an accessible price. Universal access is no longer a moat.

Foundation models, priced per token, available to all

So far, AI mostly works on paper.

Production is where 95% of pilots die. Crossing that gap takes an engineer few companies have. That engineer is the product.

In production
Stalled in pilot
Bought, unused
MIT NANDA, State of AI in Business 2025 · schematic

The frontier is shipping the engineer.

Palantir ships forward deployed engineers with its products, and OpenAI and Anthropic hire the same role. The applied AI lab is the second model, built on the frontier. Augusta Labs proved it in Europe.

01

Business

How the company makes money, where it leaks time, and who decides.

02

Product

What to build, what to skip, and what people will use.

03

Technology

Which loops need a person, and which the system can carry.

Most firms employ all three skills. Almost none employ them in one person. Pitanga Labs hires one engineer with all three.

The service

Every engagement has three stages.

01

Map

Inside the business, finding where the system pays.

02

Prove

One scoped workflow, ROI in weeks.

03

Run

Expand function by function until the system runs the department.

The first win is one workflow, mapped, proven, and left running.

The window

The model works everywhere. LATAM has nobody.

+85%
Palantir revenue growth, one year
The company that named the FDE
US$ 4B
Committed to OpenAI’s deployment company
Anthropic and Google hire FDEs too
€50M
Augusta Labs valuation, two years in
The applied AI lab model · Distyl at US$ 1.8B
US$ 10B
TAM in Brazil, still unclaimed
Brazil, bottom-up estimate
Palantir Q1 FY26 · OpenAI, May 2026 · ECO, Jun 2026

Everyone else sells hours. Pitanga Labs sells outcomes.

The fieldPitanga Labs
ModelHours, projects, ticketsOutcomes on named processes
DeliveryAugments teams, hands overEmbedded, kept running
PricingBillable hoursFee on results
RiskNo one owns deliveryPitanga Labs carries it
ReuseEach project from scratchBuilt once, redeployed
Named processes, named owner, fee on results.

Porto Alegre trains world-class engineers.

The city has Brazil’s second-best university for computer science. Its graduates leave for remote US jobs, offers abroad, and São Paulo.

Guilherme Oliveira · Founder

One of those graduates is coming home.

UFRGS computer science, then a decade building in Germany and the US.

Founding engineer to CTOBuilt embedded lender Banxware from zero to 30-plus engineers, through Series B.
Backed by UniCreditOne of Europe’s largest banks led the Series A, then invested again. €25M of €40M raised.
AI against real moneyUnderwriting, document and SRE agents in production at a regulated lender, not in a demo.
Fortune 500 deliveryGap, Walmart and European scale-ups. Built mission-critical systems for consumer peak loads.
Both sides of the tableBoard meetings, fundraising rounds, investor due diligence, from the operator’s chair.
Previous roles · CTO, principal engineer, senior engineer

The kernel companies run on. Pitanga Labs builds it.

The engineer who built the system stays until it runs the work.

Guilherme Oliveira [email protected] pitangalabs.com
Let’s talk