Introducing Waldo, the geospatial data agent Learn More

Introducing Waldo, the AI agent that accelerates the development of solutions for the physical world

Waldo, the spatial data agent in now available in public preview. Being capable of solving complex geospatial data engineering and analytics tasks, Waldo is the AI agent that accelerates the development of solutions for the physical world. Waldo can be used at no additional cost and is enabled in all plans, including the free trial of the Professional Edition. Get started now.

Waldo is also on Product Hunt. Find us there to claim $250 in extra Wherobots credits on top of your free trial.

Why we built Waldo

The world has problems that business and society need solutions for. The impact of wildfires, floods, and rising temperatures is scaling in the wrong direction. Energy, food, and service distribution isn’t effectively delivered. New infrastructure is developed where resources won’t last. And the costs of goods, customer attention, and transportation is increasing.

The data that can be used to drive real progress against these challenges exists. But the limiting factors are access to the talent who can turn that data into solutions, and efficient tools and AI designed to transform this data economically.

We built Waldo so more teams can do this work, and so the experts who already do it can do more of it. And we gave it access to tooling optimized for inferring and processing spatial data at any scale.

As an AI agent, Waldo is the crafty spatial data expert that’s ready to join your team, takes on as much work as you can give it, and it knows how to deploy Wherobots’ spatial-first tooling to complete the task at hand.

We believe AI agents like Waldo can help make the world and businesses better, and we are excited to see what you build with it.

How Waldo works

AI agents are great at code generation and answering questions about a map! But productive work with physical-world data involves more than just generating code, and the role of a spatial data agent should be more than a Q&A machine.

As a complete spatial data agent, Waldo automates the workflow that engineers and analysts would otherwise perform by hand, accelerating the delivery of complex data products and analyses from raw imagery, embeddings, raster, vector, and tabular datasets. This means companies can deploy use cases (and maps!) with spatial data faster.

Waldo accelerates the spatial data workflow: a cycle of selecting datasets, wrangling formats, performing spatial reasoning, running inference and analytics, building maps, and delivering data products, starting from a natural language prompt

Starting from a natural language prompt, like “Build me a dataset and map of parcels in Phoenix AZ that have solar panels”, Waldo iterates towards a working production data pipeline, result, or visualization to achieve the objective.

Waldo determines which datasets are useful, extracts the key spatial relationships, decides if computer vision models are required to achieve an objective, and iterates with the user to generate and test production-quality inference and analytics code that runs inline, at-scale. Along the way, it builds maps so you can inspect intermediate results and share insights. When the work is done, Waldo delivers the resulting data products into your existing Iceberg-compatible data platforms and can run the work on a schedule.

Here are example data engineering prompts that Waldo can build out today:

  • Compute mean NDVI from Sentinel-2 for every field in the Fields of the World sample over Iowa, then map the 100 lowest-vigor fields.
  • Detect [land change type] around our [infrastructure sites] in Texas whenever new Sentinel-2 imagery arrives, and update the infra_change dataset in the [my_AWS Glue_Data_Catalog_table].
  • On a schedule, process the latest GPS data from [fleet_position dataset] and map trips that are off-route or behind schedule. Flag trips likely affected by weather or traffic and drivers more than 10% over the speed limit, and write the results to my [Databricks catalog].

If you want a map or a specific answer from your data, try this:

  • Create a surface-water catchment geometry for the West Boulder River drainage in Montana with a digital elevation raster, and map this with the latest NAIP imagery as a base layer. Create monthly rainfall, snowpack, and snowfall statistics for the extent of the drainage.

Give Waldo natural language prompts like these, and Waldo takes care of the heavy lifting using the products we optimized for physical-world data.

Waldo carries decades of our team’s spatial expertise with every session. That expertise comes from research, from operating and scaling production pipelines in Wherobots Cloud, and from leading the evolution of spatial support in Apache Spark, Apache Flink, Apache Iceberg, and Apache Sedona which our team created and actively leads.

See Waldo work on a real problem. Join us on December 15 at 10 AM PT for a live session. Submit your own physical-world problem when you register, and we’ll run a few of them through Waldo live. Register for the live session.

Waldo’s optimized tooling

Waldo acts like a Formula 1 driver of Wherobots’ spatial-first products, using each in the development workflow.

  • The Havasu Catalog makes it easy to get started because it already hosts many of the raster and vector datasets (50 and growing) you may need. It provides the most spatial data type and format support of any catalog, and enables interoperability with Iceberg tables on S3 as well as foreign catalog systems like Databricks Unity Catalog and AWS Glue Data Catalog. These integrations allow you to securely bring your private data into analytical scope without having to move it.
  • RasterFlow ingests, prepares, performs inference, and ships derivatives from imagery as a pipeline using hosted computer vision models or models you import.
  • WherobotsDB provides scalable and efficient ingestion, transformation, and raster and vector data processing.
  • Wherobots Maps enables customers to inspect Waldo’s work and create interactive visualizations across multiple spatial data formats, from Zarr to Parquet to PMTiles.
Waldo brings capability and AI to your data: file formats and data integrations on the left, Waldo in the center, and the Wherobots platform products it uses on the right

These products give Waldo broader coverage for the spatial data workflow than you’ll find in other AI agents. We are rapidly expanding support for various raw and derivative data types, formats, and storage systems to expand Waldo’s capabilities and meet the needs of customers. We are also making it more likely to support your use case upfront, by hosting more open and partner datasets in the Havasu Catalog. Waldo improves with every improvement in the platform, and we want to know what you’d like support for next.

Security

We designed Waldo with model training, privacy, and data isolation in mind. You can learn more by reading our security and privacy documentation for Waldo.

Integrate and go

Wherobots supports an open lakehouse architecture. This choice gives you freedom to modernize, and seamlessly leverage Wherobots as an extension of your existing data platform to solve spatial data engineering and analytics problems for your organization.

  • Wherobots interoperates with Databricks Unity Catalog, AWS Glue Data Catalog, and Amazon S3.
  • The Havasu Catalog now exposes an Iceberg REST endpoint, so engines such as Snowflake, DuckDB and SedonaDB can query your Wherobots tables, including Iceberg V3 geometry columns, without copying data.

Considerations

Waldo:

  • is not designed to be an “ask a question, get an answer about the planet” agent. While it can be used for this type of task and can answer questions using data or a map, a result can take time due to the reasoning, iterations, and data processing required.
  • is for spatial data work, and if it doesn’t think it’s being used that way it will tell you.
  • best operates when it is integrated with datasets in S3 registered in the Havasu Catalog, or registered as foreign catalogs in Databricks Unity Catalog or the AWS Glue Data Catalog. It can help you source datasets from the web if needed.
  • is continually being optimized to improve effectiveness and time-to-workflow completion. Quality is Waldo’s KPI we are managing internally.
  • has token limits to prevent abuse. Waldo offers up to 55 million tokens during the free trial, and 140 million tokens per week for paying customers. You can reach out to us if you need more.
  • is in preview, and we are looking for your feedback to make it more useful for your use case. Reach out to product@wherobots.com and share your feedback. You will hear from a person if your request is legitimate, we promise.

Waldo Tobler and the First Law of Geography

Waldo Tobler was a geographer who stated the First Law of Geography in 1970: “Everything is related to everything else, but near things are more related than distant things.”

Tobler’s law describes the work Waldo does. Most spatial data analyses exist to describe how things of interest relate with other things. Where homes sit in proximity to flammable vegetation and fire stations is an important analysis for insurers. How a truck’s GPS trace relates to road grade, a storm cell, a delivery window, and traffic are important for fleet operators. Advertisers care about the relationship between a billboard’s possible or actual placement, real estate costs, footfall, demographics, and vehicle traffic. Waldo does the hard spatial data engineering and analytics work for teams that need to understand these types of relationships over space and time.

Start today, at no cost

The Wherobots Professional Edition includes a generous free tier of Waldo tokens and a 14 day free trial of Wherobots. Sign up, give Waldo your first prompt, and let it help you validate your idea about the physical world.