How Slingshot Aerospace’s Agatha Tool Harnesses AI to Make Sense of Orbital Activity

As space becomes more congested and contested, the space industry and governments look to more advanced solutions to make sense of on-orbit activity and be aware of potential bad actors. The 2024 Satellite Technology of the Year winner Agatha AI was born out of a collaboration between Slingshot Aerospace and the Defense Advanced Research Projects Agency (DARPA) to analyze megaconstellations.

When Slingshot Aerospace CEO Tim Solms accepted the award, he paid tribute to the team that developed the technology. An 18-year U.S. Army veteran who worked in hardware and software at Dell, Microsoft, and VMware, Solms says that while he doesn’t write code, he saw the “pent up value” in Slingshot’s trove of data, and what could be unlocked by processing that data with AI.

In this interview, Solms talks about how Agatha AI is being put to use to secure space operations, and the company’s focus on talent and culture that allows Slingshot to punch above its weight in technology.

VIA SATELLITE: Congratulations on Slingshot’s Technology of the Year Win. What did it mean to you for the company to win for Agatha AI?

Solms: Let me recognize CEO ego and hubris and put those aside. I don't write code. I do not build LLMs — yet I'm the person that gets to stand up and receive the award. There is a huge part of impostor syndrome that comes up — How do I get to do this? I represent the company, but how does that feel for the seven people that put their heart and soul into building Agatha? I wanted to make sure that I called them out. It’s not my work, but it’s my responsibility.

VIA SATELLITE: What is the story of how Agatha AI was developed by Slingshot and DARPA?

Solms: I’m going to start a little bit further back. We have three co-founders — [two of the three] Melanie Stricklan, who is an iconic figure in this industry, and Thomas Ashman, were wearing uniforms in the U.S. Air Force, which is now the U.S. Space Force, and saw that space was exceeding the Department of Defense (DoD) capability.

The DoD was reaching out to the typical systems integrators to try to fill the gaps and they did not think that that was going to be sufficient. They decided to step away from their very successful careers and start Slingshot, based on demonstrated national security and Department of Defense needs. It was not aspirational. [The need] was there and has continued to scale.

From the early days of Slingshot, DARPA said there is something unique in analyzing megaconstellations. There was alignment between what the team had already been working on and what DARPA was asking for, so they developed Agatha. The program at DARPA was called PRECOG [Predictive Reporting and Enhanced Constellation Objective Guide]. When we looked at Agatha, we thought this wildly surpasses even the highest expectations of what we thought of what we thought this AI tool was going to be able to do.

VIA SATELLITE: How do you describe the capabilities of Agatha AI?

Solms: Agatha, in the simplest of terms, identifies anomalies in activity and constellations. We can look at a commercial constellation and the team can say that of your 3,000 satellites, this one satellite has two thrust motors that are inoperable. This really happened. And the customers say ‘How in the world do you know that?’ It’s very clear to see how that behavior is different.

When we start looking past the congested and the complex and we look at the contested side of things — we know that space is an active warfighting domain. There are some bad actors in space and they tend to hide some of their activities in large constellations. Agatha is very, very good at identifying that. We don't attribute intent. We're not an intelligence agency, but we can say what we see happening. We also have contextual intelligence that we build in the model that says we know what it is and we know what it's supposed to be doing.

VIA SATELLITE: How is the tool able to get to the level of detail to make a recommendation that a thruster is not operating correctly?

Solms: I talked about contextual intelligence. We did an acquisition of Seradata which is the ‘one of one’ contextual database for all objects in space, going back to Sputnik. Everything down to the subcomponent level. That informs our AI models, and it reduces what I call the ‘sphere of uncertainty.’ We know more about an object, even studying supply chains, who’s buying what. If we don't know about a launch, we can attribute things to it once it shows up in orbit.

We study the behaviors of objects. If there are 10 different types of objects in this constellation, we know what they are, and this one is not doing what it's expected to do. Agatha says the way it's maneuvering is different, and that tells us that two thrust motors are probably inoperable.

Agatha can look at the GPS constellation and tell in near real-time, any spoofing or jamming that is being directed at the GPS constellation. If you're dependent upon GPS, you know if the signal is getting jammed because your GPS signal becomes unreliable. But what about spoofing? What if that signal is being deliberately manipulated so that the user is getting bad data? Agatha can detect that and tell you where it's coming from. We’ve found that to be very powerful.

VIA SATELLITE: What does this capability mean for your customers?

Solms: Situational awareness is absolutely critical. Even if you're not operating in a megaconstellation, you need to know what is going on around you. Whether you're a commercial entity or governmental, you're still threatened either by the complexity of the congested nature of orbital operations or by deliberate, nefarious activity in a highly contested environment.

Our sensors are designed in a way that we are persistently watching. We [collect] about 8.5 to 10 million observations per day based on optical ground sensors that are all over the world. That persistent watching sees something anomalous and we use AI as we're curating that data to make that observation meaningful and useful to the person who's receiving it — whether they are government or commercial.

VIA SATELLITE: How do you see greater opportunity for Agatha to be used by the government?

Solms: When you go back to Mel Stricklan and Thomas Ashman, they wisely predicted a spike in orbital operations that needed capability, and that's what we're doing. We see trends happening. Where we are now, we know adversarial first moves will happen in space. We look at space as red and blue, and tradecraft tells us it's probably the gray objects that are going to be the real threat. We track red and blue very closely. But gray is where we see people hiding — objects not widely tracked, not widely watched, not widely reported upon. We've constructed our sensors, our analytics, and our AI to watch that and say we see something that needs to be identified and reported immediately. We are dedicated to supporting the mission of the United States and United States allies, and we want to do everything that we can to support that.

VIA SATELLITE: Can you share an example of some anomalous on-orbit behavior that Agatha has played a part in identifying?

Solms: I'm going to use China for example. China has megaconstellations, and they like to fly their satellites in strings of three, there’s triplets all over a megaconstellation.

One of the things that we saw in analyzing was that everything was behaving normally, except in one triplet, there were two objects that maneuvered like twins. They would move up, they would move down, and they would move up, they would move down. Agatha immediately said — there's an anomaly here. We think it's somebody hiding activity in a megaconstellation.

VIA SATELLITE: What would you like to see change in how the government works with commercial companies like Slingshot?

Solms: We put out an open letter [to the U.S. government] that it is prioritizing space, but not funding it to support prioritization. A congressman from the House Armed Services Committee [recently asked me] ‘What is your biggest issue right now?’

Contracting officers’ desks are empty. They were all working remotely and took the check and left. It takes five to seven years to grow and certify a government contracting officer. It’s not that things are slowed down. They're told an award has been made but the contract isn’t going to go anywhere because there is literally no one to process it. You’ve asked for a capability. The industry is addressing that need, and you have no way to get to it.

VIA SATELLITE: Where is that situation headed?

Solms: I think the industry is going to force consolidation and I think consolidation is going to limit optionality for the government customer. If we don't fix contracting, if we don't fix a few other things, then there will be consolidation. Instead of having 11 companies to bid [there could be] three. That's not good, we want a crowded market. The government wants and needs options. They need us innovating and competing.

VIA SATELLITE: What do you think has set Slingshot Aerospace apart in developing Agatha AI?

Solms: First, I'd say it's our customers. We have a customer set that has allowed us to be nimble and dynamic. But the secret sauce in my mind is the data. We have been able to create the largest corpus of commercially available space data in existence. The data attracts the talent. We're punching way above our weight class in PhDs. I ask them all the time, ‘Why are you here? How do I keep you here?’

They say [they] get to do things here that wildly exceed what [they] can do in a research environment, more data to model and play with. Data scientists, astrodynamicists, they love coming here. My job is to keep it well capitalized, to ensure that environment stays in place. Ultimately, my job is to drive culture. Culture is the number one thing. You can have all the data in the world, but if the culture isn't there to empower it, cultivate it, and curate it, then ultimately they are going to go somewhere else. We want to make sure that we keep allowing people to do things that are interesting and disruptive and make sure they are empowered to make an impact in their roles.

VIA SATELLITE: Finally — With AI as such a hot topic right now, how do you think about meaningful AI capabilities versus AI as a buzzword?

Solms: It’s funny. It was cloud, and all the way back to dot com. The thing I love about data — it's not a bubble. Data continues to scale and be used in different ways. AI is a way to find faster, more elegant insights that come from data.

What’s unique is how that delivers a capability to a customer. Like I said, I’ve never written a line of code in my life. The market tells me whether our AI is differentiated or not. We have that feedback across four or five completely unique buckets. When I did my diligence on the company, I knew it was there, but I did not know that it was as good as it is. Now I’m starting to see market differentiation because of it.

As a CEO, I need to protect that. I have a great chief of staff who went to Atlanta [recently] to meet with two people from that team just to take them to lunch and make sure that they were connected. The first thing I asked is ‘How are they? What can I do?’ Because great AI talent is going to be pursued. You are only as good as your talent. We’re a 140-person wonky space company. How do we punch against our weight in AI? We create a culture that incentivizes people to stay and be a part of what we're doing. It starts with the guy who gets to stand up and receive the trophy saying, ‘This is not my work. Thank you to those teams.’

VIA SATELLITE: When I interviewed Melanie Stricklan for the 2022 Executive of the Year Award, she said very similar things about the importance of culture.

Solms: She's amazing, we talk all the time. We both come from the military. In the military, [when] there is a change of command, the flag goes from the past commander to the next commander. We talk about the change of command, but I am super keen on our founders’ story. There is a power to a founder story, if you're willing to honor and respect it. I think we do at Slingshot, and I want to continue to do so. VS

Lead photo: A network diagram of Slingshot Agatha data depicting the relationships among satellites in multiple, real-world, foreign constellations. Photo: Slingshot Aerospace