Between Two Quarters: Sara Dwyer on How Parambil Is Winning the Highest-Stakes Cases in Legal AI

Guest hosted by Skylar Dorosin, Principal at nvp capital

For this installment of our Between Two Quarters Founder Series, Skylar Dorosin sat down with Sara Dwyer, Founder and CEO of Parambil, an nvp capital portfolio company, to talk about the origin story behind the business, how she landed and kept her earliest customers, and what she's learned scaling a company while raising two kids under two.

Parambil is an AI platform built for complex litigation: the high-dollar, high-stakes cases that go to trial. The company takes in the mountains of documents attorneys accumulate throughout a case and turns them into the insights, chronologies, and case theories trial teams need to win.

Skylar and Sara first met over coffee, and it didn't take long for Sara's intensity to make an impression. Skylar remembers being struck by her deep background in the underlying data, the sheer pull she was already getting from early customers, and a conviction about the problem that was obvious within minutes of sitting down. That first meeting is a big part of why nvp capital invested early, and why we've loved watching Sara build ever since.

The Founder Story: From Healthcare Data to Anchor Boxes

Sara is an engineer by training. At 19, she joined a small group inside McKinsey focused on AI and life sciences, years before "AI" meant what it means today, and spent the better part of a decade buried in enormous healthcare data sets, helping patients tell their stories through their own medical records.

Parambil's first idea wasn't Parambil at all. Sara and her now co-founder, Dr. Ralph Horwitz, originally set out to help doctors unlock patient data directly. It didn't work: the models were too expensive at the time, and the incentive structures on the provider side weren't there to support it.

The real insight came from Ralph's own experience as an expert witness. In legal cases, he'd be handed "banker boxes," stuffed with PDF medical records, and asked to reconstruct a patient's story with no ability to go back and ask questions or order new tests. Unlike medicine, law doesn't let you gather more data. You work with the box you're given.

When ChatGPT launched in late 2022, Sara and Ralph moved fast. They believed this specific, messy, high-stakes data set, years of medical records tied to a single legal case, was about to be transformed by generative AI, and they built Parambil to be the company that did it.

Landing the First Customer: Rejected, Then Undeterred

The early days were, in Sara's words, humbling. Parambil's first prospective client, a law firm in Pennsylvania, backed out a week before launch. The lead partner didn't want AI anywhere near his courtroom.

So Sara went back to basics: if someone would take a meeting, she'd get on a plane. Her actual first customer came from a cold LinkedIn message to the CEO of a plaintiff's firm in Chicago, written, as she describes it, in the voice of "what would a really confident male engineer founder say." The message sold a two-person team as an "A+ team. He responded in under four hours and gave her a meeting date.

Sara showed up in Chicago with a menu of the biggest problems she'd heard firms in the space were facing and asked him to pick one to test her on. He did. That meeting became the foundation of the company's first real customer relationship, and the model Parambil still uses today: show up with a specific, credible hypothesis about the customer's problem, and ask them to pressure-test it live.

Building Trust with a Skeptical, High-Stakes Buyer

Neither Sara nor Ralph came from law, which she now counts as an advantage: it kept the team's egos low, kept them curious, and forced them to actually listen rather than assume they already understood the space.

Trust, in a business built on the facts that determine whether someone gets compensated for what happened to them, isn't optional. Early on, the team tried to earn it by making every output flawless. Sara, Ralph, and co-founder Liam Gordon would stay up manually checking every citation the product generated before it reached a client.

That turned out to be the wrong instinct. Perfect outputs handed to a client without visibility into how they were produced didn't build the client's trust; it just built the team's confidence in their own work. The shift that mattered was giving attorneys the ability to see and drive the reasoning themselves, so they could decide where they did and didn't trust the AI's output. That change, from "trust this" to "see how we got here," reshaped how the product was built from that point forward.

On the human side, trust was built the old-fashioned way: showing up. For six weeks, Sara flew to Chicago every week to sit with that first customer, sleeping on a mattress on the floor of her brother-in-law's apartment, listening and building alongside the firm in real time.

It's a version of a pattern we've heard from other founders on this show. Roshan Patel of Arrow talked about walking door-to-door into doctors' offices with a dozen donuts in hand, because in a market where the buyer is invisible to digital channels, distribution doesn't come from better tooling, it comes from physical presence and doing things that don't scale. Sara's weekly flights to Chicago are the legal-industry version of the same insight: in high-trust, high-stakes markets, you can't earn credibility from a distance. You have to show up, in person, over and over, until the trust is real.

Scaling Through Force, Not Just Intention

Growth didn't always follow a plan. In January 2025, around the same time Parambil signed its largest contract to date, supporting one of the country's biggest mass tort matters across more than 50 law firms, the team of three doubled in size. A month later, Sara had her first child and went out on leave.

What followed, in her words, was "messy and magical." Sales and engineering sat in the same 600-square-foot room, and every customer call fed directly into product changes, sometimes to Sara's own private frustration, but always in service of a fast feedback loop that let the company move at the speed its customers needed.

The real signal that Parambil had built something real came when attorneys started using the product in ways the team never designed for, building creative workflows of their own and coming back to describe what they'd figured out how to do with it.

Why Complexity Is the Moat

Most legal AI startups went after motor vehicle cases: lower value, higher volume, easier to automate, faster to monetize. Parambil deliberately didn't. Instead, the team spent more engineering time and more customer time on the hardest problems, like ingesting and reasoning over 100,000-page medical record sets, that competitors were skipping.

That bet is now the company's differentiation. Parambil is the only platform that can read fetal monitoring strips, a critical data source in birth injury cases, against ACOG clinical guidelines. When prospects who already use other AI tools run a complex case through Parambil, they see firsthand how specialized the product is for exactly the cases that matter most to their bottom line.

It's also why the "hard to build, hard to destroy" framing resonates: in a moment when many AI products are easy to stand up and just as easy to displace, Parambil's moat is the years spent solving problems that require both engineering depth and genuine domain expertise.

Pricing and ROI: It's Not About Saving Time

Parambil's pricing has moved through several phases: high when models were expensive, lowered as costs came down, then adjusted again as usage climbed. Today, the company charges per case, which works well for its market because firms can pass that cost through to clients as a standard case cost, rather than needing a dedicated AI budget line.

The more interesting shift is how customers measure return. It isn't about shaving a few days off filing a complaint. As Sara put it, the real ROI is turning a two-million-dollar settlement into a much larger one. One customer in Connecticut used Parambil throughout a case that resulted in the largest verdict in the state's history, a result the firm said would not have been possible without it.

From Blockers to Believers: Bringing Experts Into the Platform

Early on, if a nurse or doctor got access to the platform, it was often a bad sign: a signal that Parambil hadn't yet built enough trust, and that those experts saw the AI as something likely to miss what only they could catch. That's flipped. Experts now push to get onto the platform because it helps them reach their own conclusions faster, and Parambil is building out an expert network to bring them more directly into the product experience for both firms and the experts themselves.

Choosing Early Partners: Long-Term Conviction Over a Quick Check

When Sara raised her seed round, she was looking for partners who'd stay engaged not just for the next year, but the next ten. She describes herself as part traditional founder (engineer, strong resume) and part atypical: no prior startup experience, and a mom of two young kids building a company at the same time.

Her advice to other founders picking early investors: look for conviction in you specifically, not just the market you're in, and look for a partner who can offer visibility into what other founders are seeing in real time. For Sara, nvp capital was one of those partners: a firm whose portfolio was full of other vertical AI founders working through similar questions, like pricing models, customer trust, and team structure, that she could learn from as she built.

Advice for Founders Building Vertical AI

Asked what she'd tell other founders trying to break into legacy industries, Sara pointed to three signals worth looking for in a market:

1. Size without saturation. Go after markets that are large but haven't already been picked over by foundational model providers.
2. A wary, relationship-driven buyer. The right customers won't dabble; they want to know you, often in person, before they trust you with anything that matters.
3. A buyer who wants a partner, not a DIY kit. The best customers want a best-in-class provider to own the problem, not a consultant to help them build their own version.
   

Her broader advice for teams navigating fast-moving technology: pick people the same way you pick software. Hire generalists who are flexible, curious, and willing to work hard, because the shape of the job itself will keep changing.

Final Thought

Sara's story is a reminder that the founders who win in hard, trust-driven industries aren't the ones with the flashiest demo. They're the ones willing to fly to Chicago, sleep on a mattress on the floor, and spend years on the un-glamorous, hard-to-automate parts of the problem that actually matter to the customer. That combination of technical depth, genuine curiosity about an industry she wasn't originally part of, and relentless follow-through is exactly what excites us about backing founders this early.

Watch the full episode here: