Meet Data Ocean
The platform that connects the Sabiá to dozens of Brazilian databases to answer with real, up-to-date, sourced numbers, in the chat and the API.
There is enormous potential to improve the lives of millions of Brazilians with accessible, well-organized information. Health signs can be recognized earlier, students can learn with proven, more effective methods and small business owners can make decisions based on concrete data, drawing on the kind of analysis that until recently was only within reach of large companies.
For the most part, that data already exists. Brazil produces high-quality public statistics about almost everything, from health and education to jobs, public safety, climate and the economy, but that information lives scattered across dozens of portals, in technical formats, full of codes and acronyms that only specialists master. For most people, in practice, it might as well not exist.
Data Ocean was born to close that gap. It is Maritaca’s platform that integrates multiple Brazilian databases, public and private, into the Sabiá, our most advanced language model. With this integration, instead of relying only on the generic knowledge recorded during training, the Sabiá queries the sources directly at the moment you ask, cross-referencing the numbers and answering with real, local context, always telling you where each figure came from.
Data Ocean in action: a question in plain language, Brazilian data queried on the spot and the answer with its source.
Where the answers come from
A traditional language model answers with what it memorized during training, knowledge that ages quickly and carries no date and no source. Web search helps with recent information, but it returns ready-made pages rather than the original data that allows deeper analysis.
With Data Ocean, when it receives a question the Sabiá decides which databases to query, writes and runs the queries, cross-references results from different sources and even computes and generates charts when the question calls for it. The entire process takes seconds, without you having to know SQL, be familiar with data portals or figure out on your own that the answer was scattered across three different agencies, because that work now belongs to the model.
The result is answers built on numbers that came from the original database, queried on the spot, and that arrive with the source cited so you can check them.
An ocean of Brazilian data
Today Data Ocean covers dozens of databases adding up to billions of records. Some of the public sources it integrates give a sense of the variety.
| Domain | Some sources | What you can find out |
|---|---|---|
| Economy and finance | Central Bank, IPEA, B3, CVM | Selic rate, inflation (IPCA), exchange rates, stock prices since 2000, investment funds, public companies’ financial statements |
| Companies | Federal Revenue (CNPJ) | Sector, size, partners and address of the country’s companies, with openings and closings |
| Jobs and income | CAGED, RAIS, PNAD Contínua | Salaries by occupation and city, jobs created and closed, unemployment |
| Health | ANVISA, CNES, DATASUS, Fiocruz | Medication package inserts, clinical protocols, health facilities and beds, vaccination, epidemics |
| Education | INEP/MEC | School Census, Higher Education Census, IDEB quality scores for schools and municipalities since 2005 |
| Population and territory | IBGE (2022 Census) | Demographics and income by neighborhood and census tract across the country |
| Public safety | FBSP, SINESP, state secretariats | Incidents by crime type, municipality and period |
| Climate and environment | INMET, INPE, ANA | Temperature and rainfall for every municipality since 1951, plus wildfires, deforestation and river levels |
| Agriculture | IBGE, CONAB, MAPA | Production by municipality, commodity prices, harvests, rural insurance |
| Foreign trade | ComexStat/MDIC | Exports and imports by product and municipality since 1997 |
| Government and elections | Transparency Portal (CGU), TSE | Public spending, transfers and budget amendments, plus candidates, results and voter profiles |
| And more | SUSEP, consumidor.gov.br, ANAC… | Insurance, consumer complaints, flights and airfares, tourism, energy, sanitation |
The full list is larger and grows every month, and it also includes private databases that complement the official statistics. Some queries happen in real time, such as the weather forecast, dengue alerts, river levels and deforestation alerts, which come from the official sources at the moment you ask.
What you can do with it
If you want to know the best neighborhoods in your city to open an ice cream shop, the Sabiá cross-references demographics, income, foot traffic and competition to guide the choice. If you need to understand the public safety picture, the job market or the quality of schools in a region, it researches, analyzes and delivers a well-built answer with real, local context. Below are a few examples of what you can ask.
“Which neighborhoods in Campinas have high income, lots of foot traffic and few ice cream shops? I want to open one.”
“I’m deciding between two cities in the interior of São Paulo. Compare safety, schools and price per square meter.”
“What are the side effects of this medication? And which health facility near me has a pediatrician?”
“Has education in my state improved in recent years? How do we compare to the national average?”
“How has rainfall behaved in my region over the last 30 years? And what happened to soybean prices in that period?”
“Compare the performance of the largest real estate funds with the CDI over the last five years.”
In every case, the Sabiá goes beyond answering questions, because it interprets the data, draws correlations and generates useful insights, always delivering the numbers along with the source so you can check, dig deeper or even disagree.
What an answer looks like
To show how it works in practice, we asked the chat for the strategy of a lawsuit over the late delivery of an apartment bought off-plan. In the example below, the Sabiá looked up what the courts have already decided about construction delays, laid out the timeline of the delay and worked out how much can be claimed for each item, based on the case’s numbers.
| Claim | Legal basis | Estimated amount |
|---|---|---|
| Inverting the 0.5% monthly penalty against the developer | STJ binding precedent 971 | ~R$ 40.3k |
| Lost rental income, as a subsidiary claim | TJSP súmula 162 | ~R$ 49k |
| Moral damages for the lengthy delay | STJ case law | R$ 20 to 30k |
| Refund of construction-phase interest paid during the delay, if any | TJSP case law | ~R$ 27k |
| Replacing the INCC index with the IPCA on the balance during the delay | TJSP case law | R$ 5 to 15k |
Recommended value of the case: between R$ 80k and R$ 130k, depending on which claims are included. The Sabiá also warned that the inverted penalty and lost rental income cannot be combined (STJ binding precedent 970), which is why the second claim comes in as subsidiary, and confirmed that the 180-day grace clause is valid (TJSP súmula 164), so the delay counts from July 2025.
Answer produced by the Sabiá with Data Ocean in July 2026, condensed from the chat. Binding precedents and súmulas located in the case-law database; amounts computed from the contract’s numbers.
The full answer also brought a suggested structure for the filing, section by section, and the table of precedents with links to cite. The legwork of research and calculation is done in minutes, and the strategy stays in the lawyer’s hands.
Harder questions
Not every question is solved with a single cross-reference. Some require combining spot prices, futures curves, exchange rates and interest rates to support a decision, combining epidemiological surveillance, mortality and vaccination records to guide public management, or rebuilding five years of price series to rethink a portfolio. We asked three more questions in the chat, with no instruction beyond the text itself, and summarized below what the Sabiá returned.
Sell the harvest or hold it
A soybean farmer’s decision depends on numbers that live in different places, such as the spot price from the CEPEA indicator, B3 futures contracts, the PTAX exchange rate and the Selic rate, which sets the cost of keeping the grain in storage. In the example below, the Sabiá gathered this information, cross-referenced it and presented the final analysis.
| Scenario | Value per 60 kg sack |
|---|---|
| Sell now, at the spot price (CEPEA/Paranaguá) | R$ 132.84 |
| Sell in July, at the B3 futures curve converted at the exchange rate | ~R$ 125 |
| Monthly cost of holding the grain (interest, storage, risk) | R$ 6 to 9 |
| Waiting until July, net of carrying costs | ~R$ 117 |
The Sabiá’s recommendation: sell most of the harvest now. The market was paying a premium of R$ 7 to R$ 11 per sack for immediate delivery, the futures curve did not cover the cost of carrying the stock and the dollar at R$ 5.14 favored revenue in reais. The suggestion was to sell 70 to 80% and hold the rest only under specific conditions, such as storage costs well below average.
Answer produced by the Sabiá with Data Ocean in June 2026, condensed from the chat. Spot price from the CEPEA/ESALQ indicator in Paranaguá, B3 futures, PTAX exchange rate from the Central Bank and the 14.25% yearly Selic rate on the date of the question.
A city’s dengue picture
Whoever runs a health network needs to keep track, all at once, of how transmission is doing right now, what happened in recent years and how much of the population is already protected, information that comes from three different databases. In the example below, the Sabiá queried all three and brought the overview together into a single answer.
| Dimension | What the data showed |
|---|---|
| Epidemiological alert | Green in the week of the question, with transmission falling (Rt of 0.83) |
| Deaths in the 2024 epidemic | 91 records, against an average of about 4 per year in the four previous years |
| Death profile | 71% of the 2024 victims were 60 or older |
| Vaccination (ages 10 to 14) | About half the target group received the first dose and 28% completed the schedule |
The Sabiá’s diagnosis: a moment of respite in a city that remains vulnerable. The recommendation was to use the low-transmission period to reach those who have not completed the vaccination schedule, prepare observation beds and train teams to manage fever in the elderly, the group that concentrated the deaths in 2024.
Answer produced by the Sabiá with Data Ocean in June 2026, condensed from the chat. Alert from InfoDengue (Fiocruz/FGV), deaths by cause and age from SIM/DATASUS, doses administered from the National Immunization Program.
Where to let the money grow
Repositioning a portfolio means looking backward and forward at the same time, rebuilding the real return of each investment over recent years and projecting what comes next. In the example below, the Sabiá pulled the monthly series of the CDI rate, inflation, the dollar, the stock index and three real estate funds since July 2021, wrote and ran the code that nets out income tax and inflation for each one, and crossed the result with the Central Bank’s Focus survey.
| Investment | Gross return | Net real |
|---|---|---|
| HGLG11, logistics real estate fund | +87.2% | +38.5% |
| Ibovespa, with dividends | +80.4% | +29.1% |
| CDI | +76.3% | +26.5% |
| VISC11, shopping-mall real estate fund | +47.6% | +13.1% |
| HFOF11, real estate fund of funds | +38.5% | +9.7% |
| US dollar | -0.9% | -23.9% |
The Sabiá’s reading: with inflation at 30.3% over the period, money left in dollars lost almost a quarter of its purchasing power, while the CDI returned 26.5% above inflation net of tax. For the next 12 months, with the Focus survey projecting the Selic at 14%, inflation at 5.3% and the exchange rate at R$ 5.20, the math put the CDI (+6.0% net real) ahead of inflation-linked Treasuries (+4.5%) and stocks (+3.0%). The final suggestion was a portfolio with 50 to 60% in fixed income, 20 to 25% in stocks, 10 to 15% in real estate funds and 5% in dollars as a hedge, with the reminder that the analysis is not financial advice.
Answer produced by the Sabiá with Data Ocean in July 2026, condensed from the chat. CDI, inflation and exchange-rate series from the Central Bank, prices and dividends from B3 and CVM filings, expectations from the Focus survey. Inflation-linked Treasuries were left out of the historical ranking because the available series was incomplete, as the Sabiá itself pointed out.
All the answers came out of the same flow: the Sabiá decided which databases to query, fetched the numbers on the spot and turned them into a recommendation you can check.
Where to try it
Data Ocean is already part of the conversation at chat.maritaca.ai, where you just ask normally and the Sabiá decides when to query the data. In the API, you turn the tool on with the data_ocean flag and the whole flow runs inside Maritaca’s infrastructure, as the post Integrated tools in the Sabiá API explains, with the full documentation available at docs.maritaca.ai.
We believe intelligence only fulfills its role when it puts itself at people’s service, and Data Ocean is our most concrete step in that direction, putting Brazil’s data one question away.