Analyzing and predicting climate and weather is one of the most important parts of the space economy. Machine learning and artificial intelligence (ML/AI) has been a key part of that for years.
A new company in Toronto called Perceptive Space, however, is taking that to another level. Having just emerged from stealth a few months ago, Perceptive Space and founder/CEO Padmashri Suresh aim to deploy ML/AI to understand and predict the weather in space itself.
SpaceQ spoke with Suresh this week about the growing issue of disruptive space weather, and what Perceptive Space is doing to address it.
Space weather a growing concern
One might be forgiven for wondering “how can there be weather in space? It’s a vacuum”. Space isn’t actually a vacuum; aside from the planets and meteors and comets and the like, we also have to contend with the “solar wind”; streams of charged particles that travel from the Sun throughout the solar system, including Earth. Sometimes the wind is comparatively calm, but sometimes it can be intense, particularly during Coronal Mass Ejections where the Sun ejects enormous amounts of radiation and charged particles.
On Earth, in most cases, we’re protected by Earth’s atmosphere and magnetosphere. The “solar wind” helps to create the Aurora Borealis that many Canadians are familiar with. The particularly intense geomagnetic storms that these ejections can cause can cause enormous damage, however, such as in the 1859 Carrington Event where the telegraph system of the time was severely affected by intense solar weather. In 1989, a solar storm caused a significant number of blackouts across Quebec as well, leaving some people without power for days. At the time, there were concerns that it may have been some kind of Cold War attack; nobody had expected or predicted the geomagnetic storm.
Now, however, the effects could be much more dire. The large geosynchronous satellites used at the time of the Quebec blackout were usually hardened against the radiation and charged particles caused by solar weather. In the twenty-first century however, we’ve moved to relying on enormous numbers of small satellites in low Earth orbit (LEO), many of which are neither designed nor capable of enduring intense solar weather and geomagnetic storms. A bad storm could have serious impacts on LEO constellations and the people that rely on them for communication, observation, and all their other twenty first century tasks.
These past few years, there have been several reminders of this vulnerability. In early 2022, a geomagnetic storm caused the loss of 38 Starlink satellites, following a coronal mass ejection. And just a few months ago, in May of 2024, NASA said that we saw a “barrage of large solar flares and coronal mass ejections“ that was responsible for “one of the strongest displays of auroras on record in the past 500 years.”
Starlink satellites are comparatively resilient, and the recent barrage apparently had no serious effect on the constellation. The growing number of small satellites means that inclement space weather could not only disrupt capabilities that terrestrial customers rely on, but potentially leave satellites in crowded orbits as unmovable debris.
And, needless to say, a coronal event could pose a serious threat to crewed space missions.
A background in both machine learning and space weather
According to Suresh, these sorts of incidents are the key reason why Perceptive Space exists. She said that “while space debris poses risks in specific regions and activities, space weather—comprising radiation and plasma effects—is a universal concern for all space-based activities, impacting both satellites and human missions.” Yet while this is a real concern, “commercial solutions have largely focused on mitigating the risks posed by space debris and space junk.”
Suresh has always had an interest in space, going back to her time as a NASA Earth & Space Science Fellow at Utah State University. She studied space weather there, and said that she “frequently encountered challenges related to space weather—radiation, atmospheric drag, and other effects that impacted our design, operation, and launch decisions.” The issue, however, was that “these decisions were often made with the understanding that space weather predictions were not very accurate.”
Prior to her time at Utah State, she worked at IBM in ML/AI, and she said “[that] got me thinking: “why couldn’t we [use ML/AI to] improve the predictability of space weather? We’ve been studying the sun and gathering data about the space environment for decades”. After stints at Los Alamos National Lab and in Washington DC, she concluded that the answer was yes, and that there was a market for more accurate space weather predictions.
So she returned to the tech industry, building up her expertise and track record in deploying ML/AI, and began laying down the foundations for Perceptive Space.
Using AI to predict space weather
Perceptive Space is a company that uses ML/AI to predict space weather, helping customers avoid and mitigate its effects. It is building a platform that customers will be able to use to predict space weather that, according to Perceptive’s site, features “space weather APIs and dashboards for LEO missions.”
Slated to officially launch in early 2025, they’ve already begun a pilot program. The company is already taking on “early adopters/pilot customers” according to Suresh, though she could not get into specifics on those early adopters.
Suresh also couldn’t reveal details on specific sources and methods, citing intellectual property (IP) security concerns, especially about their “bleeding edge ML/AI”. She did say that the company uses “data collected about the Sun and the space environment by both ground and space-based instruments”, however, as well as “data about operations of technologies in space” for ML/AI training.
The idea, Suresh said, is to “use ML/AI in tandem with traditional physics-based approaches” to “extract more signal from the same data.” The algorithm can notice patterns that humans can’t, and flag potential problems that humans would miss, leading to “better decision-making capabilities for assessing impact of space weather. She also noted that they’re taking an eclectic approach, saying that as Perceptive Space is “an applied ML/AI company, we are not restricted to developing or using one technique or aspect of these technologies.”
This may include the red-hot generative AI technologies that are attracting so much attention on Wall Street and Silicon Valley, though Suresh did not specifically mention generative AI.
A “critical mass” and the emergence of Perceptive Space
The company was originally formed in 2022, and emerged from stealth a few months ago in 2024. Those dates are striking; 2022 was when the Starlink satellites were knocked out by space weather, and May of 2024 had what NASA called the “barrage” of flares. Was Perceptive Space’s entrance and exit from stealth a reaction to those two events, or coincidental?
In our interview, Suresh said that the answer was “in some ways, yes.” She said that she had “wanted to build a company around this for a while,” especially as the space economy grew by leaps and bounds over the last decade. She recognized, though, that there needed to be a “critical mass”: a space economy large enough for space weather to prove to be an issue, and an event that signified the issue’s salience. “The loss of [the 34] Starlink satellites seemed to be that critical mass,” she said, and so she “ended up talking to hundreds of space and defense companies” to understand the pain point and how her company could address it. After those conversations, Perceptive Space’s time in stealth began.
As to the 2024 events, Suresh said that they’re relevant, but less directly salient. She said that they were representative of how “we are currently in the midst of a solar maxima,” and therefore “will continue experiencing space weather events like geomagnetic storms and solar flares more frequently.” This could prove to be a serious problem for a lot of young companies in the space economy, many of which “haven’t faced the brunt of space weather” before, and who now have greater awareness of the problem. That said, she was clear that the May storms were “not necessarily influencing our direction in any new way.”
A $3.9M investment in a “distributed team”
The emergence from stealth came with an announcement of funding: an oversubscribed pre-seed round of $3.9M CAD, which included investments by Panache Ventures, Metaplanet, 7Percent Ventures, Mythos Ventures and AIN Ventures. Suresh said that as a company bridging the gap between ML/AI and the space economy, they attracted interest from “strategic investors who are generalists, AI-focused investors and space (defense) investors to invest in the company.” An impressive raise, in a difficult environment for fundraising.
Aside from building the platform, the main goal for the raise is to grow the team. Suresh said that Perceptive has a “highly technical team…[everyone] on our team has an aerospace (or space physics) and machine learning background.” The focus has been on people with skills relevant to “understanding the physics of the space weather, engineering of the satellites, and how to build scalable machine learning systems.”
When asked about hiring for such demanding skill sets, Suresh said that “there’s a lot of excitement about AI and Space, and rightfully so because of their role in shaping humanity’s future”, and so “being a space company that is leveraging AI has certainly helped in attracting talent.” Their other key advantage for attracting talent, though, is that while Perceptive Space is based in Toronto, they’re a “distributed team”. Perceptive Space is remote-first, with team members in Toronto, Oakville, Los Angeles, and (even Norway. Their job advertisements on LinkedIn emphasize the remote work aspect, and have advertisements for both Canadian and US-based talent. Suresh said that “this helps us hire the best talent anywhere in the US or Canada”.
In an era where many other software companies are pushing skilled workers to return to the office, requiring some to deal with disruptive moves across the country to cities with much higher housing costs, this could be a welcome change for key talent in the sector. It could prove to be a critical advantage in the search for talent, as Perceptive Space continues building their platform for launch next year.