Applied-science AI lab · Latin America
Most AI predicts. We build the layer that turns prediction into the right decision — calibrated, constrained, and accountable. Frontier science, made to be acted on.
We build AI that decides — not predictions that win benchmarks, but decisions that hold up when you act on them.
Everyone can predict. Almost no one decides well.
The last decade made prediction cheap. Any team can now forecast demand, score risk, or flag an anomaly. Yet the hardest part of any operation was never the forecast — it's the decision that comes after it: what to do, under real constraints, real costs, and real uncertainty, when being wrong is expensive.
That gap — between a model's output and a defensible decision — is where value leaks out of most AI projects. It's the gap we close.
From signal to decision.
We work across three connected layers. They're not three products — they're one pipeline.
See what's coming, honestly.
Forecasts, risk, and time-to-event models built for calibration and robustness — not leaderboard scores. The number means something, and it holds up when the world shifts.
Make your models agree.
Real organizations have many models that disagree. We fuse them — machine learning, statistical, and structural — into one coherent, trustworthy picture. This is our signature work, and it's harder than it sounds.
Turn it into the best action.
We translate predictions into the decision that's actually optimal under your constraints, your risk appetite, and your cost of being wrong. This is where prediction finally pays for itself.
We're not racing to build the biggest model.
The world has enough giant general-purpose models, and more capital chasing them than we could ever match. We chose the layer they ignore: the decision layer.
Think of it this way — the large models are the brain. We build the judgment: the system that takes any model's prediction and turns it into the best possible decision. That's where industrial decisions are actually made, where the data is yours and no one else's, and where world-class work doesn't require a billion-dollar compute bill.
Most AI is graded on accuracy. We grade ourselves on whether the decision was right — a different, harder, and far more useful target. Models that win benchmarks routinely lose money. Models built around the decision don't.
Every line of research we pursue has to clear one bar: it must be both publishable as science and deployable as a product. If a result can't improve a real decision, it isn't ours.
Hard decisions, high stakes.
We focus where decisions are expensive, repeated, and rich in data.
Dispatch, contracting, and planning under uncertainty.
Capacity, risk, and resource allocation.
Capital allocation and risk under competing outcomes.
Forecasting and operations where small decisions compound.
Don't see your sector? If the decision is hard and the data is real, talk to us.
A lab, not an agency.
Datistica is built on a simple conviction: the fastest path from frontier science to real-world value is to do both under one roof. We publish, we open-source, and we ship — and each one makes the others stronger.
Peer-reviewed papers and open benchmarks, not slideware. Our research is public so our work can be judged on its merits.
Research that stays in a PDF helps no one. We build the deployed systems that carry it into the world.
Every project sharpens a reusable capability. The lab gets smarter with each problem we solve.
Read our work.
We believe in showing, not claiming. Our research is where you can see how we think — and where our latest results live.
Publications coming soon — follow us on LinkedIn.
A lab built for scientists.
Latin America's best researchers shouldn't have to choose between doing real science and being treated fairly for it.
We're building the place we wish existed: where you work on frontier problems with real data, keep your name on what you publish, and own a real piece of what you create — without surrendering your ideas to keep doing the work you love.
Do frontier science. Keep your name on it. Own a piece of what you build. Stay a scientist.
What we're building.
Datistica is Latin America's home for applied decision science — a lab built on the conviction that the region's data and talent deserve frontier-grade work done here, not imported. We're research-first and product-bound: we publish, we open-source, and we ship, because the shortest path from a new idea to a real decision runs through all three.
Have a decision worth getting right?
Whether you're an organization with a hard problem, a researcher who wants to build, or a partner who wants in early — we'd like to hear from you.
We read everything. We reply to what fits.