About

About São Tomás

A climate risk company born in the crisis room: rain and flooding, wildfire, asset protection. We build platforms for people who decide with the weather in front of them — and we hold firm opinions about how a number reaches the screen.

Who we are

São Tomás builds and operates climate intelligence platforms for operations whose assets are exposed to the sky. We serve utilities, the environment sector, agriculture and forestry, civil defence, infrastructure — highways, railways, ports —, telecom, water and sanitation, mining, logistics, governments, and the financial and insurance market.

Before we were a product company, we were a team that responded to real crises — the Rio Grande do Sul disaster, the 2024 wildfire crisis, operations with civil defence. That is not biography: it is what determines which questions the platforms answer.

And there was a turn worth telling, because it explains the rest of this site: we stopped selling climate and started selling management supported by climate. We do not hand a coffee grower a forecast — we hand them the management of their field’s windows: when to plant, the best day to dry, which window in the next ten days to harvest. Climate stopped being the product and became the input to a business decision. It is the difference between selling a thermometer and selling health.

We deliver projects, not shelf licences. A climate risk platform is only useful when it speaks the language of the operation using it — the activities, the thresholds, the safety constraints and that sector’s calendar. That is why scoping comes before the contract, and the demonstration runs on a real slice of your operation.

Who builds it

Climate science, remote sensing and software engineering in the same team — and it is that combination that lets us own the models instead of reselling somebody else’s forecast.

25+ years: of senior technical experience
ITA: and Environmental Physics in the team’s background
CBERS: the Brazilian space programme, in the résumé
4: proprietary products in operation

The partners

Four partners who have worked together since 2016 — the team has been stable for almost a decade, long before "climate tech" became a category. This is not a team assembled for an investment thesis.

Vitor Baccarin Zanetti

Co-founder · Product Owner, Risk360

PhD in Sciences from ITA and MSc in Environmental Physics from UFMT. Conceived and led the development of all four proprietary products. Led a team of 25 in the national response to the Rio Grande do Sul disaster and in supporting the fire brigades through the 2024 wildfire crisis. Researcher on the Brasil 3 Tempos project (SAE/PR · UNDP).

Luiz Carlos Silva Filho

Co-founder · AI and digital transformation

MSc in Artificial Intelligence, over 23 years in C-level leadership and more than 50 large-scale projects across government, health, energy, climate, telecom and finance. Delivered enterprise platforms serving tens of millions of users.

Carlos Eduardo Toledo de Miranda

Co-founder · Data science

25 years in research, development and innovation, with pioneering work in orbital remote sensing and geospatial Big Data — including the CBERS programme, the China–Brazil Earth Resources Satellite.

Roberto Mizuuti

Co-founder · Engineering

Over 25 years in software engineering, from architecture to operations. Responsible for the platforms’ architecture, for the computational implementation of the models, and for AI engineering in production — the person who makes what the science proposes stand up and run at scale.

Where we have been

Before we were a product company, we were a team in the crisis room. That is where what we build comes from.

Power utilities

Real climate risk cases over transmission assets in operation.

Rio Grande do Sul

National response to the climate disaster.

2024 wildfire crisis

Supporting the fire brigades through the emergency.

São Paulo civil defence

Joint monitoring and response operations.

Santos Climate Action Plan

Full involvement in the city’s plan, from start to finish.

ITV-Vale

Installation, configuration and tuning of the Climate Forecast climate model, with technical advisory to the project.

GIZ

For the German development agency, a landslide risk and environmental vulnerability study in Salvador, Bahia.

Iepé

With the Institute for Indigenous Research and Training, training young Indigenous leaders in environmental and territorial management in the Tumucumaque Park Indigenous Land.

Clients

Agribusiness, engineering, geospatial technology and international organisations — the same decision machinery, in contexts that look nothing alike.

Technology partners

We do not rebuild what already exists and works. The geospatial foundation, the sensors and the cloud come from the people who do it best — and that is what our models run on.

How we treat numbers

These are engineering rules, not marketing ones — they live inside the platforms’ code, and they are why our results survive an audit.

AI explains, the engine decides

In regulated domains, risk classification is deterministic, versioned and auditable. AI agents write the technical and executive readings and guide the decision — without ever classifying the risk. That is the architecture that survives the auditor.

No number on screen is invented

If the source does not publish a field, it stays absent — not zeroed, not blank, not derived from a plausible formula. Zero looks honest and is a trap: “0 mm accumulated” reads as “it did not rain”, when the fact was “nobody measured”.

Uncertainty is part of the number

A projection without a percentile band and without model agreement is half the information. We would rather ship the band, even when it is wide — especially when it is wide.

Every disclosable number carries its origin

Dataset, calibration, how many models were used, how many cells. Without provenance, a number does not survive the auditor’s second question.

Source coverage is declared

When an asset falls outside the footprint of the source feeding it, the platform says so — instead of silently displaying demonstration data.

Alerts follow the operator’s criteria

When the client already has defined thresholds — the same ones their civil defence uses — the alert follows those, not ours. Two systems raising different alarms about the same rainfall is worse than no alarm at all.