In 2017, three entrepreneurs — Chris Hazard, Mike Resnick and Mike Capps — came together to launch a platform for building AI and machine learning tools geared toward the enterprise. Hazard and Resnick had been working on various AI and game projects for the U.S. military, while Capps had recently retired as president of Epic Games. The aforementioned platform eventually became Diveplane, which today offers products that create synthetic data to train AI systems, find anomalies in data and forecast market trends.
In a sign that business is healthy, Raleigh, North Carolina-based Diveplane today closed a $25 million Series A funding round led by the defense-focused fund Shield Capital, with Calibrate Ventures, L3Harris Technologies and Sigma Defense participating. Capps tells TechCrunch that the new capital will be used primarily to grow the company’s roughly 20-person headcount and create new internal departments, starting with customer success.
“We built a platform that’s [easy] to use, faster, transparent, auditable and explainable … We provide the tools to developers and data scientists, who manage data input and output entirely on their hardware or cloud,” Capps said in an email interview. “We’d love to see our tools in as many hands as possible, and that’s a big part of the reason for this fundraise.”
Diveplane’s technology, which Capps claims has a history in government agencies including the U.S. Transportation Command, spun out of Hazardous Software, the company Hazard founded after working as a software architect at Amazon-owned Kiva Systems and Motorola. Capps met Hazard through a mutual acquaintance, and they collaborated with Resnick to develop a Diveplane proof-of-concept.
Diveplane occupies the MLOps category of AI startups, which aims to furnish organizations with tools to deploy and maintain machine learning models in production. For example, the company’s Geminai product creates anonymized, statistically similar “twin” datasets to train AI systems in a putatively privacy-preserving way. (Training on synthetic data has its downsides, it’s worth noting.) Diveplane’s Sonar service, meanwhile, performs a regular analysis of data and AI systems to ensure that the systems don’t drift off course — i.e. become less accurate in their predictions — over time.
“Our tech works with messy data, sparse data and small data sets … [a]nd our unique single-model approach means you train once for any sort of task, so you can follow the signal in your data,” Capps explained. “[I]t’s all editable [and] online, so when you need more or different data, or find bad data that needs removal, you can change on the fly without starting from scratch. If a prediction doesn’t look right, you can trace exactly what training data influenced the prediction. And it’s all auditable over the lifetime of the model, so you can roll back to the state of the system, re-create a classification and then pull the full explanation for it.”
On the synthetic data side, Diveplane competes with startups like MostlyAI, Gretel and Hazy. And in MLOps more broadly, it goes head to head with rivals such as Arize, Tecton and Weights & Biases, the last of which raised $135 million last October.
To stand out, Diveplane has focused a portion of its customer acquisition efforts on defense outfits — reflecting its co-founders’ backgrounds (Capps once taught at a Naval post-graduate school, and Hazard worked for the Department of Defense).
As L3Harris’ Dan Gittsovich put it via email: “DoD customers are intensely focused on responsible use of AI to improve their decision speed when lives are on the line, and we are confident Diveplane’s explainable trusted AI solutions could help combatant commanders now. We’ve also seen that Diveplane’s unique and powerful toolkit makes AI application intuitive for nearly any user, so we believe it will provide a discriminating advantage for both campaign planning and urgent operational missions.”
Capps described the rest of Diveplane’s customer base as “larger enterprises.” Recently, the startup announced deals with Scanbuy and Mutua Madrileña, one of the largest insurers in Spain.
“[The] pandemic added some headwinds for us,” Capps admitted. “We were actively selling into healthcare enterprises, and the pandemic brought much higher priorities for those clients and pushed innovation to the back burner. That’s changed, and now we see increased spending combined with a widespread push for data privacy … Regulation’s certainly a tailwind for a company like ours, because the rules tend to focus on privacy, transparency and responsibility and that’s the whole reason we exist!”