In the beginning, Jay Madheswaran wanted to be a chemist. Or perhaps a materials scientist. Then, as college approached, he noticed that computer science moved at a pace that made the other disciplines feel stately. Software offered the intoxicating possibility that an idea could become a working thing before the thought had cooled. He changed direction. The decision would carry him from Facebook before its IPO to the first engineering chair at Rubrik, then across the table into venture capital, and finally back to a terminal as a founder.
The logos make a neat résumé. The more useful story is what he took from each room. Facebook supplied a close view of speed. Rubrik taught him how a product and a culture grow together. Lightspeed Venture Partners trained him to inspect a market from above. Eve, the legal AI company he co-founded with former Rubrik colleagues Matt Noe and David Zeng, is where those perspectives were folded into one specific wager: plaintiff lawyers should spend less of their day feeding the machinery of a case.
A builder takes the investor's chair
Madheswaran was born in Bangalore and moved to the United States at 10. He earned a computer science degree at the University of Illinois Urbana-Champaign, then joined Facebook while the company was still private. His next move put him much closer to zero. At Rubrik, he became the first engineer and eventually led product engineering, helping the team expand from a handful of people to hundreds.
That kind of early seat is an education with no syllabus. A founding engineer cannot retreat into code alone. Product decisions become hiring decisions; hiring decisions become culture; culture appears in the product when nobody is looking. Madheswaran later described building Rubrik from the ground up as one of the defining growth experiences of his life. He also emphasized mentorship and culture, the quiet systems that decide whether a growing organization remains coherent.
In 2018, Lightspeed recruited him from its own portfolio company and made him a partner. He focused on enterprise infrastructure, open-source software and cloud companies. On paper, he had moved from making products to judging the people who make them. In practice, the builder kept leaking through. He created sample applications to test ideas he was evaluating. The investor was still prototyping.
The paperwork behind the drama
Law is full of cinematic moments. A witness changes the temperature of a room. A final argument clicks. A verdict lands. The daily work is less photogenic: intake calls, medical records, chronologies, discovery responses, document review, drafts, checks, follow-ups and the relentless need to know what happened next. Madheswaran and his co-founders saw an industry whose judgment work was trailed by an enormous administrative wake.
Their first idea in 2020 was broad: use large language models to help businesses make sense of the documents that define their work. Legal practice offered complexity, abundant text and obvious day-to-day friction. Eve emerged publicly in October 2023 with $14 million in seed funding and a platform meant to assist from intake through resolution. The company later narrowed its attention around plaintiff-side firms.
That focus carries an economic insight. Plaintiff lawyers often work on contingency. They earn a share of a recovery rather than billing for each hour. When software compresses a day of document work into minutes, their revenue model does not lose a billable unit. The firm gains room to examine another matter, call another client or think harder about the case already open. Madheswaran has called these firms the canary in the coal mine for professional services because their incentives make the value of saved time unusually plain.
Capital followed the focus
Disclosed funding rounds, drawn to relative scale
A $47 million Series A arrived in January 2025. In September, a $103 million Series B led by Spark Capital valued Eve above $1 billion and brought disclosed funding to $164 million. By April 2026, the company said more than 1,000 plaintiff firms were using Eve across more than 200,000 active cases. Those are company-reported figures, useful as a picture of the direction and speed rather than a substitute for the details of any individual firm's experience.
Growth also changed the job in front of Madheswaran. A founder who once had to explain why a lawyer might use generative AI now leads a company selling into firms that already arrive with opinions, experiments and a crowded software drawer. The question has shifted from whether AI belongs in legal practice to which parts of the practice should depend on it. That raises the standard. A clever demonstration may win a meeting; it cannot carry a case for months. The product has to survive messy records, house styles, permissions, deadlines and the ordinary resistance of people whose existing process mostly works. Madheswaran's answer has been to move closer to the whole workflow, even as doing so makes the promise harder to keep.
The trust problem wears a suit
Legal software has a special burden. A writing tool can be fluent and wrong in the same breath. A confident error wrapped in legal syntax may be more dangerous than a clumsy one because it looks ready to file. Madheswaran's early guidance to firms was stubbornly practical: inspect security and retention, ask how the product handles hallucinations, demand a fit with the firm's real workflows, and make adoption tangible with a recently closed case.
Eve was designed to encourage citations and direct quotations, with lawyers prompted to validate the output. The balance matters. A machine can read every page without fatigue, but it has never watched a jury decide which detail feels human. It can draft a demand, but it cannot accept professional responsibility for sending it. The more capable the system becomes, the more consequential the reviewer's attention becomes.
Madheswaran is frank about where the bottleneck goes. When AI accelerates production, review becomes scarce. He has seen the same dynamic in software engineering: code can be produced faster, leaving experienced humans with a growing queue of decisions to inspect. The constraint has moved uphill. This is progress, provided the organization redesigns around it instead of pretending the constraint evaporated.
From assistant to operating system
Eve's product arc mirrors that larger ambition. The early platform helped with tasks such as case evaluation, research, medical record review, chronologies, demand letters and discovery. Eve 2.0, introduced in January 2026, added proactive agents, an auditor and an analyst. By June, the company renamed the broader system EveOS and described layers for case data, agents, auditing and analytics.
The shift is architectural. A useful assistant waits for a prompt. An operating system attempts to connect what happens before and after that prompt: the call that creates a matter, the records that arrive later, the deadline hiding in a document and the management question about why cases stall. Madheswaran's pitch grew from faster tasks to a firm that can see itself.
There is a founder's loop inside that evolution. At Facebook, he saw rapid iteration at scale. At Rubrik, he learned that enterprise products must survive both technical complexity and human organizations. At Lightspeed, he learned to look for a painful wedge and a team able to execute. Eve's wedge was document-heavy casework. Its expansion is the belief that once software earns a place in the matter, it can help coordinate the firm.
The useful form of ambition
Madheswaran often frames the aim in terms of empowering plaintiff firms against better-resourced opponents. The rhetoric is large; the mechanism is wonderfully mundane. Read the stack. Find the inconsistency. Draft the chronology. Flag the deadline. Show the citation. Give the lawyer enough time to make the call that cannot be delegated.
His career has crossed engineering, operations and investing, but it keeps returning to the same appetite: turn an abstract possibility into a working system. The chemist he nearly became might appreciate the method. Combine a clear incentive, a repetitive burden and software that improves quickly. Keep professional judgment outside the reaction vessel. Then watch where the pressure moves.
For Eve, the next test will not be whether AI can produce more legal text. It plainly can. The test is whether firms become more attentive as production becomes cheaper: more attentive to evidence, clients, strategy and the quality of the final human review. Madheswaran has built his company around the wager that removing paperwork creates room for judgment. The room is open now. Lawyers still have to decide what to do with it.
Keep reading and watching