Launching today

Waldo by Wherobots
The AI agent that accelerates engineering with spatial data
259 followers
The AI agent that accelerates engineering with spatial data
259 followers
Waldo is the geospatial data agent that turns questions about the physical world into answers, maps, and production-ready data pipelines. It analyzes satellite imagery, geometries, geography data, and your organization's first party data, writes and debugs the analysis, and delivers results in hours. When the work needs to run continuously, Waldo turns it into a scheduled pipeline on your existing data platform. Built for spatial analytics at scale by the creators of Apache Sedona.













Free Options
Launch Team / Built With

AWS Startup AdvisorPersonalized AWS guidance that meets you where you’re already building
Promoted



Hey Product Hunt, and happy GIS Day! 🌎 I'm Jia, co-founder of Wherobots and one of the original creators of Apache Sedona.
Today we're launching Waldo in public preview, and for once, Waldo is easy to find 🔍
Waldo is the AI agent that accelerates development with physical world data: the spatial data expert your data team always wished it had. It takes spatial data engineering and analytics work from backlog to shipped, in hours instead of weeks.
😩 The problem
Spatial data grows by petabytes every day, and the experts who can use it are scarce. A head of data with one spatial engineer and twenty open requests ends up declining eighteen.
General AI coding agents get spatial relationships wrong and stop at the code.
Legacy GIS and general cloud data platforms struggle at regional to planetary scale, and imagery work gets skipped.
🌍 How Waldo works
Ask in plain English
Waldo finds the right open and private datasets and checks them for fit
It writes and runs the analysis and imagery inference, and builds maps so you can inspect each step
You stay in the loop the whole way
It ships a pipeline, dataset or map to the Iceberg-compatible platform you already run (Databricks, Snowflake, AWS Glue), on a schedule if you need one
💬 Try a question like
"Which homes in El Dorado County sit on high wildfire-risk terrain and are more than a 10-minute drive from the nearest fire station?"
🔎 Why "Waldo"?
Named for geographer Waldo Tobler and his First Law of Geography: "Everything is related to everything else, but near things are more related than distant things." That's the work Waldo does, so launching on GIS Day felt right.
🎁 Try it today
Waldo tokens are included at no extra cost during the preview; you pay only for compute. Sign up at wherobots.com
You found Waldo! Claim $250 in extra credits with this form (mention Product Hunt)
What spatial data project is stuck in your backlog? Drop it below, and we'll build it with Waldo and share what we get 👇
Is it possible to explore large datasets without writing complex SQL?
@daniel_harris11 Yes, and that question is why we built Waldo.
WherobotsDB and RasterFlow are our engines for planet-scale data processing and AI inference on Earth imagery, with Python and SQL as first-class interfaces. Our users love them, but until now every analysis started with writing a lot of code.
With Waldo, you ask in plain English. Waldo writes the Wherobots code, runs it on those same engines, and hands back maps, dashboards, and a report in a notebook you can share with colleagues. You can also schedule any analysis as a recurring task, so you see what changed in the physical world every day.
Here is a screenshot of Waldo writing the Python and SQL code that runs our AI models to detect rooftop solar panels across the US:
Can Waldo help businesses make better decisions by finding useful patterns in their location based data?
@angelo_bram Yes! You can ask a business question in plain English: “Which of our stores have the most competitors within a mile?” or “Which delivery zones have the most buildings we don’t serve yet?” Waldo finds the data it needs and works out the answer.
Take a retailer picking its next store. Waldo pulls sales by location, nearby competitors, and building density, joins them on WherobotsDB, and ranks the neighborhoods where demand is high and competition is thin.
The same approach works for an insurer sizing up exposure or a logistics team choosing a warehouse site too!
@angelo_bram Yes!
For instance, Waldo can support the geospatial risk-analysis component of a property and casualty underwriting digital assistant. The assistant could screen properties against hazard layers, identify concentrations of insured value, analyze claims by location, and flag conditions that meet approved referral rules. These capabilities require suitable property, exposure, claims, and hazard data.
A prototype could use Waldo for conversational exploration and spatial queries. A production application could connect through the Spatial SQL API and SDKs, with an underwriter interface, approved rules, and sourced explanations.
Would you recommend Waldo for teams just getting started with spatial AI?
@imogen_wallace 100%
If a data / AI engineer has no experience with Spatial AI, Waldo can be a great point to start. In a few minutes, you can run inference on large-scale earth observation data using state-of-the-art spatial AI models (Tile2Net, SAM3, CanopyHeights...), and even productionize the output within within your daily work. You won't need to orchestrate GPUs or anything of that sort. Waldo will automatically invoke RasterFlow to run inference at scale on the spatial data you select, and then uses WherobotsDB to push the results in your AI repo or in the format you pick. Give it a whirl
@imogen_wallace Yes Adding one reason to Mo's answer: Waldo shows its work, so a new team learns spatial analysis as it goes!
Your team can skip memorizing 300+ spatial functions. Name the place, the measure, and the output in plain English. If the prompt is ambiguous, Waldo asks up to three clarifying questions, each with a default.
Every answer lists its steps under "Worked": the code, the run time, the rows read, and the datasets and dates behind the result. Copy that code into a notebook when you are ready to build on it. Waldo asks for approval before it starts a runtime or a billed task.
The trial runs 14 days with no credit card, $95 in compute credits, and 55 million Waldo tokens.
How much time can Waldo save when working on geospatial projects?
@barnaby_lloyd Depends on the project, but in our experience most of the time goes to setup, before any analysis happens 😭: finding the right datasets, loading them, figuring out how they relate, then writing and debugging the code.
Waldo does all of that for you. It searches your catalogs (Databricks Unity Catalog and AWS Glue too), pulls in public datasets you don’t have yet, writes and runs the Python or SQL, and puts the result on a map.
And the compute is fast. Everything runs on WherobotsDB and RasterFlow, the fastest and most scalable spatial compute engines out there, so it works the same for one city or the whole planet.
Waldo is in the free trial, so the easiest way to find out is to bring one of your own questions and time it. Docs: https://docs.wherobots.com/develop/waldo/overview
geospatial analysis can get complicated pretty quickly. Having an agent that writes and debugs the analysis could save teams a lot of time. How easy is it to get started?
@lily_hart1 You can sign up on the Wherobots website and all compute is fully provisioned and managed for you on AWS. The free trial lasts 14 days and includes $95 in compute and no cost for tokens during the initial trial period for Waldo. If you mention that you signed up on Product Hunt on the Waldo page (there is a form at a the bottom) we will issue you another $250 in credits for your trial.
@lily_hart1 Once you sign up, your first answer is one prompt away.
There is nothing to install and no data to load. Waldo runs in the browser, and Wherobots Open Data is already in your catalog, including 2.5 billion Overture building footprints and US Census boundaries.
A good first prompt from our quickstart: "Which homes across Mesa, Arizona have rooftop solar and which don't?" RasterFlow detects the panels in aerial imagery, and WherobotsDB joins them to building footprints. You approve the compute, then get a summary, a table, a map, and the code behind it.
Quickstart: https://docs.wherobots.com/get-started/waldo-quickstart
Congrats @drjiayu How well does it perform when datasets become really large?
@drjiayu @luz_bidelspach Thanks for the question! Waldo runs on Wherobots, which is built for distributed compute. Queries in Wherobots can scale to tens of even hundreds of compute nodes, easily handling very large scale or complex datasets. One customers is the Overture Maps Foundation, and we enable overture to process the planetary scale buildings data pipeline as much as 20X faster than their original architecture. We provide all of Overture data in the Havasu catalog which Waldo also has access to.
@luz_bidelspach @benppruden
Adding numbers to Ben’s answer. Waldo runs large jobs on two engines.
WherobotsDB runs the analytics on vector and raster data. It completed all 12 SpatialBench queries at scale factor 1000, including heavy spatial joins the next best Spark-based engine with Spatial SQL could not finish in 10 hours. It ran a Texas-statewide raster spatial join in 3 minutes 28 seconds. GeoPostcodes joins population rasters with postal boundaries worldwide, and that refresh went from 39 days to under one day.
RasterFlow runs AI models on imagery. One of its largest deployments processed 348 TB of satellite imagery, and model runs across the continental US and five other countries each finished in minutes to a few hours.
Diagram below: how Waldo combines both in one pipeline