FX Leduc

François-Xavier Leduc runs a company most people would find dull on paper. DeepIP builds artificial intelligence for patent professionals - the attorneys and in-house teams who draft, file, and defend the documents that protect inventions. It is precise, technical, deadline-bound work, and it has barely changed since the 1990s. Leduc looked at that and saw the opportunity of his career.

DeepIP, which he co-founded in 2024 and leads as CEO from New York, is a workflow-native AI platform meant to sit inside the tools patent teams already use. Rather than a standalone chatbot bolted on the side, it threads through the full lifecycle of a patent, from the first spark of an invention through prosecution, enforcement, and portfolio management. That design choice is the whole thesis. "The first wave of AI in patent practice focused on speeding up individual tasks," Leduc has said. "But patent work is cumulative." Each step builds on the last, so an assistant that only helps with one task leaves most of the value on the table.

The market agreed quickly. In early 2026 DeepIP raised a $25 million Series B co-led by Korelya Capital and Serena, with Balderton and Headline joining, bringing total funding to $40 million. Revenue had grown tenfold over the previous 18 months. The customer roster reads like a who's who of the field: Greenberg Traurig, Mewburn Ellis, Cantor Colburn, Schwegman, along with corporate innovators like Philips and Dexcom. In June 2026 the company acquired Munich-based PatentMaker, a signal that Leduc intends to consolidate a fragmented market rather than wait it out.

The pattern behind the ventures

If DeepIP feels like a departure, the shape of it is familiar. Leduc has spent his career on the unglamorous plumbing of applied AI - the parts that decide whether a smart model actually works in the real world. His previous company, Kili Technology, was built on the same instinct.

He founded Kili in 2018 with Edouard d'Archimbaud, who had been building one of Europe's most advanced AI labs at BNP Paribas. Their conviction ran against the fashion of the moment: that AI success depends less on model sophistication and more on the quality of the data feeding it. Kili turned raw enterprise data into clean, annotated datasets ready for machine learning. The company launched its product in 2020, won renewals almost immediately, and raised more than $30 million in 2021 from Serena, Headline, and Balderton - the same investors who would later back DeepIP. Its customers included Airbus, SAP, Safran, Louis Vuitton, Capgemini, Michelin, Stellantis, and Crédit Agricole.

The partnership with d'Archimbaud is itself part of the story. DeepIP is the third company the two have started together, after Doppio Software in 2008 and Kili in 2018. Leduc handles the commercial engine - go-to-market, positioning, fundraising - while d'Archimbaud, as CTO, runs the technology. Few co-founder relationships survive one company. Theirs has survived three.

Selling to skeptics

What Leduc brings to deep tech is a merchant's instinct. His background is not in engineering or law. He studied in the classical French Khâgne humanities track, took a Master in Contemporary History at the University of Paris I: Panthéon-Sorbonne, and an MSc in Management and Economics at emlyon business school. Before startups he worked at Procter & Gamble, Saint-Gobain Weber, and as a consultant for The Boston Consulting Group, where he worked on the digital transformation of a major carmaker. In 2013 he founded Tripndrive, a car-sharing company, his first turn as a founder.

That commercial fluency shows in how DeepIP found its footing. Product-market fit, by Leduc's account, arrived almost instantly - attorneys told him they would buy the tool with their own money if their firms would not. The pitch works because it is grounded in a concrete promise: give patent professionals back roughly two hours of their day, and let them produce higher-quality applications while they are at it. DeepIP reports 20% higher adoption and 40% higher usage than standalone AI tools, the dividend of building into the workflow rather than beside it.

The bet on augmentation

Leduc has planted his flag on one side of the AI debate. In a field where much of the noise is about automation replacing knowledge workers, his message is that the expert stays central and the machine does the heavy lifting around them. "We augment, not replace" is not a slogan he uses lightly; it is the design principle that shapes what DeepIP builds and how patent professionals are asked to use it. The wager is that the durable AI businesses will be the ones that make skilled people better rather than promising to do without them.

There is a through-line across all of it. Kili was about the quality of the data that goes into AI. DeepIP is about the quality of the expert work that comes out of it. Both companies exist because Leduc keeps noticing the same thing others overlook: the real bottleneck is rarely the model. It is everything around the model - the messy inputs, the human workflows, the accumulated context - and that is where he keeps choosing to build.

For now, the ambition is clear enough. Leduc wants DeepIP to become the standard infrastructure for intellectual property in an economy being reshaped by AI - the connective layer that patent teams everywhere run on. He has done the hard part before: taking a technical idea and turning it into a company that large, cautious enterprises will actually pay for. On his third try, with the same partner and the same investors, he is betting the third time compounds the first two.