Daniel Kornev knows what it's like to build an AI product people actually use. At Yandex, he helped ship Alice, a voice assistant that eventually commanded 60 million daily users—a scale most Silicon Valley founders can only dream about. By the time he founded Sentius, he'd moved on from his work at Yandex and his involvement with DeepPavlov, chasing a vision he couldn't shake: autonomous agents capable of handling the kind of soul-crushing enterprise workflows most humans would happily delegate to a machine.
That March, his new company, Sentius, secured a spot in Techstars Columbus' inaugural cohort. Thirteen weeks of intense mentorship and product refinement culminated on May 29, 2024, when Kornev pitched to a room full of investors and corporate partners at Demo Day. Nearly two years later, Sentius finds itself navigating a reality familiar to countless accelerator graduates—solid technology, some early partnerships, and a market that has become, perhaps more than anyone anticipated, brutally crowded.
The Credentials Were Never in Doubt
The founding team arrived with the kind of research pedigree that venture capitalists typically underline twice. Mikhail Burtsev, who runs DeepPavlov and holds an AI fellowship at London Institute for Mathematical Sciences, signed on as co-founder. Dilyara Baymurzina, who'd led engineering on DeepPavlov Dream and worked on the team's Alexa Prize efforts, joined as well. Artem Sagaida took the founding CTO role, responsible for the technical architecture.
They established headquarters in Belmont, California, with engineering support in Yerevan, Armenia. The legal entity: Generative Assistants, Inc.
Techstars Columbus, launched in partnership with The Ohio State University in September 2023, offered them a 13-week sprint focused on AI, quantum computing, and deep tech applications across foundational industries. Managing Director Tim Grace ran a cohort of twelve companies, most tackling challenges in healthcare, energy, or automation. Standard Techstars terms at the time: up to $120,000 for roughly 6-9% equity, with an optional $100,000 convertible note on offer.
What happened next? Nothing publicly disclosed, at least in terms of follow-on funding.
The UI Map Gambit
Most AI agent platforms lean heavily on large language models and call it a day. Sentius, when it unveiled its Teach & Repeat Platform on December 16, 2024, claimed something different. The system relies on what the company calls "UI maps"—essentially, a structured way for agents to navigate enterprise software interfaces with more reliability and, theoretically, less hallucination risk than pure LLM approaches.
The workflow demonstration sounds straightforward enough: a user walks through a process once. The agent watches, captures the sequence, builds a map of the UI elements involved, then tries to replicate the task on its own. The pitch centers on audit trails, cost predictability, and compliance-friendly operations—particularly appealing for finance teams wrestling with accounts receivable forecasting or generating customer proposals.
That same day, Sentius published benchmark results claiming a 94% success rate on WebVoyager, a dataset of 50 web-based navigation tasks. They tested their browser agent on a random sample and posted methodology notes on their research blog. At the time, 94% sounded competitive.
Whether it aged well is another question.
When Benchmarks Start to Blur
In February 2025, Sentius announced a partnership with Koop, a compliance automation platform. Together, they developed "Housekeeper," an AI agent designed to navigate the labyrinth of SOC 2, ISO 27001, and HIPAA requirements. Koop markets the product, though Sentius' involvement tends to surface in press releases rather than front-and-center on Koop's site.
It's the kind of enterprise integration that matters more than any benchmark score. Real customers, real workflows, real revenue potential.
But the competitive landscape didn't stand still. Some industry observers have noted that browser agents have begun clustering around higher scores on WebVoyager, raising questions about whether the benchmark retains any meaningful signal. Some wonder if teams are optimizing for the test itself rather than generalizable capability. Either way, Sentius' December 2024 score of 94% no longer looks as distinctive as it once did.
YC's Summer 2024 batch, presented just months after Sentius' Techstars Demo Day, overflowed with AI agent startups targeting enterprise workflows. Every accelerator cohort since has featured at least one team building autonomous process automation. The market didn't just expand—it flooded.
The Advisor Bench and the Headcount Question
Sentius assembled an advisory roster that reads like a who's who of enterprise AI and productivity software. Eugene Izhikevich, co-founder and chairman of Brain Corp, advises on business strategy. Brett Brewer, formerly VP of Microsoft 365, covers the productivity angle. Bill MacCartney, who led machine learning at Cohere and directed AI initiatives at Apple, brings LLM expertise. Anastasia Paushkina, with stints at UiPath, Microsoft, and Amazon, rounds out the robotics process automation perspective.
Access to that caliber of advisor doesn't guarantee traction, though. It never has.
Available data suggests the company operates with a small team, likely in the range of 2 to 10 employees. Recent Techstars job board activity echoed that range. For a company founded in 2023, the headcount suggests either measured growth, capital constraints, or a deliberate choice to stay lean while validating product-market fit. Possibly all three.
No subsequent fundraising announcements have appeared. Unverified third-party revenue estimates—always unreliable, often wildly so—place Sentius around $855,000 in annual revenue with a valuation near $2.8 million, apparently with no external funding beyond the accelerator round. These figures lack methodology and independent verification, but they sketch a picture of a company still searching for the business model that will unlock scale.
The Long Game in a Short-Attention Market
The Techstars Columbus program continues its six-month cadence, hosting Demo Days that showcase AI, energy, and healthcare innovations to Ohio investors and corporate partners. Sentius graduated into a different world than the one that greeted earlier AI infrastructure companies. Foundation models commoditized faster than most anticipated. Agent frameworks proliferated. Enterprises, after a parade of overhyped automation tools, grew skeptical of vendor claims.
Sentius still offers private beta access through its site. The Koop partnership demonstrates some level of validation—enough, at least, to co-develop a compliance product and bring it to market. The advisors remain engaged. The research pedigree hasn't disappeared.
But nearly two years past Demo Day, the question facing Sentius isn't whether the team can build sophisticated agents. Kornev and his co-founders have already proven they can do that. The harder question, the one that keeps founders up at night, is whether they can convince enterprises to bet on yet another automation platform when dozens of well-funded competitors are making the same promises, often louder and with deeper pockets.
Sometimes the technology works. Sometimes the timing doesn't.
