Finding every example of a feature across satellite images can take specialist knowledge, careful labeling and time. Blackshark.ai’s Orca Huntr aims to make that search easier: users mark what they want to find, and the system builds a model to look for similar objects or features.
Search begins with a mark on the image
Orca Huntr’s interface is built around a simple action. A user scribbles over an example of a target feature, and the AI searches for other instances in the imagery. A second color can mark areas that should not count, and advanced controls offer additional adjustments.
The search can stay within images a user uploads or expand across the planet. The examples described by the company include identifying buildings with solar panels, outlining burned areas in wildfire zones and counting fishing boats in an area of ocean. The same approach could also be used to locate adversary assets in a secure setting.
The idea is to make geospatial intelligence more accessible to people who can recognize a feature but may not have coding experience or a background in machine learning. Users can refine a model by adding or adjusting marks and see its response as they work.
People remain part of the AI process
Blackshark.ai says Orca Huntr is not designed to remove human judgment from detection. Instead, the user helps guide the system and improve its results. That matters because AI models are not perfectly accurate, and a person may need to correct what the model has found.
In a conventional workflow, object detection can require specialists to label many images, then train a model on those examples. The source describes that process as slow and costly at scale. Sensitive work, such as military intelligence, can also require the work to be done internally by a team with the right expertise.
With Orca Huntr, refinement can be more immediate. The source gives the example of a jet-detection algorithm that is only 80% accurate: instead of sending another hundred annotated images to a model maker and waiting, a user can add a brush stroke and see whether the model improves.
Blackshark.ai CEO Michael Putz described this interaction as a way for users to learn how training and annotation affect the results. The company sees the act of refining a model as approachable and interactive, rather than as a process hidden behind specialist services.
From image search to real-world planning
The tool could help a range of customers answer a basic question: where are the examples of a particular thing? Realtors, developers, governments, scientists and military users may all have reason to search imagery for specific features.
Some customers may want a broader service than the search tool alone. Putz said a wind farm developer could use it to identify promising areas for turbines, then ask Blackshark.ai to consider further constraints and create line-of-sight visualizations. The company described placing a wind park in a 3D environment and showing how it would appear from a town square or terrace.
Governments are another potential market, including agencies responsible for forests and coasts. Although the interface is intended to be simple, organizations may still need help applying data and AI tools to their work. The source notes that government investments in AI and data over the last decade have had mixed results, with some tools proving too expensive or limited.
A digital Earth foundation and new funding
Blackshark.ai grew out of the gaming industry and previously built a digital twin of Earth. Its work on that project involved creating a system that could interpret imagery from different sources and time periods. The company says that experience helped it build technology capable of handling varied datasets.
The company also has access to investor Maxar’s archive of satellite data, which spans different eras and satellite types. That archive can support model training and add context to other datasets, according to the source.
Orca Huntr may eventually prove useful beyond satellite imagery. Putz pointed to possible applications in radiology, where doctors could mark tumors, and said a hard drive manufacturer had made an interesting request. The tool’s underlying idea—creating models by marking examples—could potentially apply to other kinds of imagery.
Blackshark.ai announced $15 million in new funding, bringing its total raised to $35 million. The new investors named in the source include In-Q-Tel (IQT), Safran, ISAI Cap Venture, Capgemini’s VC Fund managed by ISAI, Einstein Industries Ventures, Interwoven Ventures (formerly ROBO Global Ventures), OurCrowd, Gaingels and OpAmp Capital, alongside existing investors Point72 Ventures, M12 Microsoft’s Venture Fund, and Maxar.
Paying customers were expected to be able to test Orca Huntr starting December 4. The company did not provide pricing details.