Compliance software is where glamour goes to fill out a form in triplicate. Thomas Stewart arrived there by the honest route: he needed it. Before Hadrius became a New York software company, Stewart and his co-founders ran an SEC-registered investment adviser. Registration brought the familiar luggage of regulated finance: recurring reviews, filings, approvals, employee oversight, recordkeeping and the low atmospheric pressure of a future audit. The tools meant to help often added another layer of work. So the founders, being technical, began automating their own program.
That modest act contained the company. The team was not staring at a fashionable blank canvas, asking where artificial intelligence might look impressive. It was staring at a crowded task list, asking which parts required professional judgment and which merely consumed professional hours. The distinction still anchors Stewart's pitch for Hadrius: software can collect the evidence, understand context and surface potential violations. A compliance officer should make the decision.
“The AI surfaces. The human judges.”Thomas Stewart's compact rule for agentic compliance
An education in systems, then paperwork
Stewart's academic combination now looks almost suspiciously well suited to the job. At the University of Virginia, he studied mathematics and economics, with a minor in physics. He spent time in UVA's High Energy Physics Laboratory and on the Putnam Math Team. Mathematics rewards structure; economics watches incentives; physics punishes the loose assumption hiding beneath a tidy model. Regulated finance supplies all three, then asks everyone to archive the email.
His early career moved between code and companies. He co-founded Bazza while at university, worked in artificial intelligence at Abeyon, and later founded Quantbase. Public career records describe both Bazza and Quantbase as acquired businesses. Quantbase is central to the Hadrius story because it put Stewart on the regulated side of the table. He was working from his own firm's administrative pain, where he had watched the tasks accumulate.
How the problem became the product
The route carries a useful founder lesson. A promising wedge is often a job whose nuisance you can describe with embarrassing precision. A vague complaint produces a vague product. Lived irritation gives you the vocabulary of the buyer, the shape of the workflow and a reliable sense of what cannot safely be automated away.
A company hidden in the queue
Stewart founded Hadrius with Allen Calderwood and Som Mohapatra. The company joined Y Combinator's Winter 2023 batch with a proposition aimed at investment managers: consolidate ongoing SEC compliance, use language models to review large volumes of material, and give compliance teams one place to manage the exceptions. Its early launch material spoke of email, Slack and SMS reviews alongside marketing approvals, filings and employee trade monitoring. The canvas widened, but the basic arrangement survived.
On the company's first anniversary, Stewart wrote that Hadrius had grown from three people to five, served more than 60 clients and managed compliance across more than $100 billion in assets. The numbers were early, but his long thank-you list was equally revealing. He named teammates, investors, advisers and friends one after another. Startup mythology prefers the lone genius because a crowd is difficult to fit on a poster. Stewart's account read more like the work itself: connected, specific and indebted.
By July 2026, Hadrius said more than 500 financial institutions and investment firms were using its platform, from single-person advisers to institutions with more than 100,000 employees. It announced a $22 million Series A led by CRV, bringing its combined seed and Series A financing to $27 million. Y Combinator, Pathlight Ventures and several industry founders participated. The capital was assigned to a larger ambition: bring the core surfaces of a compliance program into one AI-native system of record.
The review economy
Stewart's timing argument begins with abundance. AI makes words, images and analysis cheaper to create. Financial firms now communicate across email, Slack, Teams, WhatsApp, Signal and AI assistants, while new trading venues keep adding activity to supervise. Creation expands quickly. Review does not. A compliance department cannot simply hire in proportion to every new message and market.
This is the overlooked economy beneath generative AI. Every surge in creation produces a corresponding market for checking, ranking, attributing and preserving. In finance, the output must also survive an examination. A useful alert needs context. A useful decision needs a trail. A useful archive needs to produce the record when somebody asks for it months later. Merely adding a chatbot to a collection of disconnected databases does not solve that architecture.
“Everyone in compliance is talking about AI. Almost nobody is talking about the data layer underneath it.”Stewart on why context comes before cleverness
Hadrius now describes six areas: marketing, communications, people, trades, branches and firm audit readiness. The company says its system reduces manual compliance work by 70 percent and saves customers more than 20 hours each week. Another published figure, a 95 percent reduction in false positives, points to a priority for overloaded reviewers. A tool that produces more alarms resembles a smoke detector installed above the toaster.
The design principle is human rather than theatrical. Compliance officers remain responsible for policy and interpretation. The machine handles scale, assembles surrounding information and pushes the likeliest issues forward. This makes the final reviewer more important, not decorative. In a sector where the evidence may later be inspected by a regulator, preserving who reviewed what, under which rule and with what conclusion is part of the product.
The company after the pivot
Hadrius also carries the less polished part of a founder story: the idea before the idea. Early investor Alexander Zorychta has written that he met Stewart and Mohapatra when they were UVA students and backed them in June 2022, before Hadrius had become the business people now recognize. Others judged the pair by the project they happened to be pursuing at the time. Zorychta watched how they decided, listened, shipped and recovered when the answer changed. They pivoted, entered Y Combinator and kept building.
That history makes Stewart's confidence in the current market more interesting. Conviction did not arrive as stubborn loyalty to a first draft. It emerged after the team discarded one direction and followed a problem it knew more intimately. The distinction is easy to miss. Persistence is not the refusal to move; sometimes it is the willingness to preserve the team while replacing the thesis.
Zorychta also described Stewart and Mohapatra returning value to the community around them: speaking with UVA students, mentoring younger founders and supporting his own move into venture investing. Stewart's anniversary post showed a similar instinct. Milestones became occasions to name the people around the company. For a business selling systems of record, there is something fitting about a founder who keeps a careful ledger of help received.
Serious software, occasional trackball
Stewart's public voice is usually direct and systems-minded. He writes about unified data, measurable noise reduction and the practical governance of AI. He also shared Hadrius's April Fools' announcement of a BlackBerry edition, praising an interface that respected the “sanctity of the trackball.” The imaginary product promised BBM alerts, a dashboard optimized for 320 by 240 pixels and approve and escalate buttons separated by four whole pixels.
The joke landed because enterprise software has a long memory and compliance software has an especially distinguished attic. Hadrius is selling modern infrastructure into an industry where replacement carries risk. Humor can acknowledge the absurdity without dismissing the seriousness. The regulated buyer does not need another prophet shouting that history has begun this quarter. That buyer needs migration, records, controls and software colleagues will actually use.
What Stewart is really building
Stewart often describes compliance as the machinery that creates trust in financial markets. It is a more interesting claim than efficiency alone. Faster marketing review is useful. Fewer false alarms are merciful. Yet the durable value is confidence that a firm can explain its behavior, apply its policy consistently and produce the record on demand.
His aspiration for Hadrius follows from that framing. The company wants to replace fragmented SEC and FINRA tools with a system broad enough to see the whole firm and disciplined enough to leave judgment with accountable people. By the end of 2026, its stated product plan extended agentic oversight across the full compliance spectrum, including trades, branches and firm-level audit readiness.
There is ambition here, but also a constraint worth keeping. Hadrius earns its case when compliance officers can stop spending focused hours assembling screenshots, sampling messages and chasing evidence through five systems. Stewart's company is a bet that the tedious middle can shrink while the human decision becomes clearer and the compliance officer remains accountable.
That began with three technical founders confronting their own administrative burden. They automated a piece, recognized the pattern and crossed the desk from regulated operator to infrastructure builder. The paperwork did not disappear. It became legible enough to turn into a company.
Continue exploring