- Sedai is using deep reinforcement learning to help enterprises go beyond automation to actual autonomy
- The idea is to free up software engineers and DevOps teams to do higher-order tasks rather than focusing on day-to-day management of microservices
- While analysts agree these tools could make life easier, they warned that relying too heavy on AI could actually backfire
It may be little known today, but software startup Sedai is angling to become the Waymo of the cloud world. The company, which just bagged $20 million in Series B funding, is storming the market with the idea that cloud compute and storage shouldn’t just be automated but autonomous.
“Automation at scale has a critical problem: It needs someone to take the decision,” Sedai CEO Suresh Mathew told Fierce. “AI is enough if there’s a human to help. [But] decision making needs more than AI.”
Enter Sedai, which has used patented deep reinforcement learning to build its software into an expert Decision Engine. The idea is to free platform and site reliability engineers from having to make thousands of tiny decisions about the microservices they manage so they can work on higher-order items – all while lowering costs and improving efficiency.
Mathew said Sedai’s software can today be applied to manage compute (think virtual machines, Kubernetes and serverless resources) across all three of the hyperscale cloud environments. It has its sights set on expanding its applicability to storage and LLMops next.
On the former front, it already offers compatibility with Databricks and is looking to tackle repositories like Snowflake EMR (electronic medical record) systems next, Mathew said.
Enabling autonomy
Sedai’s ambitions for autonomy aren’t theoretical.
Mathew noted the idea for Sedai stemmed from work he did at PayPal, where he helped the company build the first autonomous underpinnings for its payments platform. When he made his exit in 2020 to found Sedai, PayPal was one of its angel investors, he said. It subsequently raised $15 million in 2022. Coupled with the $20 million it just scored, that means it's raised $38.8 million to date.
While it hasn’t really focused on its go-to-market strategy until now, Sedai has already bagged some big fish. Mathew rattled off a list of prominent customers in the cybersecurity, pharmaceutical and banking industries that have already deployed its technology. It also has a proof-of-concept in progress with one of the top three U.S. telecom companies, he said.
“The most difficult customers have already been cracked open. We are ready to soak all these verticals in,” he said.
Mathew said the money the company just raised will go in part towards R&D in the aforementioned areas but also to scaling its go-to-market functions.
Analyst angle
Asked about the competitive landscape, Mathew said it was hard to compare what Sedai is doing with what others offer the same way it’s hard to compare Waymo to Uber or Lyft.
But Gartner VP Analyst Manjunath Bhat told Fierce the company is up against the likes of Akamas, Avesha, CAST AI, CloudBolt (formerly Stormforge) and PerfectScale (which is part of doit).
Bhat added that some of the key hurdles that have thus far prevented enterprises from adopting autonomous technology include a lack of maturity in DevOps, platforming engineering and SRE; a lack of collaboration between software engineering and IT teams; regulatory restrictions; and clunky old architectures.
In terms of where Sedai can help, Bhat said the company's tools offer the most benefit when it comes to predicting demand and autoscaling accordingly; optimizing latency in serverless environments; autoconfiguring application runtimes, platforms and virtual machines via traffic analysis; and “rightsizing” recommendations to downsize or shut off instances to save costs.
But while AI and autonomous tools are certainly poised to help software engineers and DevOps teams, it's worth noting that Gartner and AvidThink's Roy Chua have also flagged a huge potential downside to relying too heavily on such capabilities.
"Although AI seeks to reduce cognitive load, it could unintentionally reduce cognitive skills as well," Bhat and his colleagues wrote in a January 2024 report. "Human cognitive skills deteriorate when not used. This leaves us unprepared to make the right decisions when AI tools reach their limits and we find ourselves suddenly back in the driver’s seat. The added risk of using AI as a decision engine is that it decreases human understanding of how systems operate and leads to a lack of situational awareness."