The story
Ulysse was founded by Perle Roche, a veterinary surgeon who spent the first part of her career in the clinical reality behind veterinary data: consultations, diagnoses, treatment pathways and the conversations with owners that decide what actually happens to an animal. A consultation note is a rich clinical document. It records the presentation, the diagnosis in the vet's own words, the bloods, the medications and the comorbidities. It is also free text, in whichever practice management system that particular clinic happens to run, and often a scanned PDF from the practice before.
The second part of her career was spent on the other side of the same industry, as a healthcare strategy consultant at ClearView Healthcare Partners in London, working on animal health commercial due diligence and performance improvement. From there the view is different. Animal health companies spend hundreds of millions each year developing new medicines, and each of the top five sponsors runs an R&D budget in the region of half a billion to a billion. Yet when it comes time to run a pivotal study, the way they find eligible patients is to ask people they know: KOL relationships, personal networks, CRO site outreach and manual chart review.
That is the gap Ulysse was started to close. Not a better clinical trial services business, and not another registry, but the missing layer in between: a system that connects to veterinary practice systems, standardises what is in the records, and lets a sponsor see where eligible animals are, then recruit them through the vet who already treats them.
Why now, and not five years ago
Three things changed at once. Practice ownership consolidated, so a single data agreement with one corporate group or one cloud practice management vendor now reaches hundreds of clinics rather than one. Language models became good enough to read free-text clinical prose and scanned histories at a cost per document that makes national coverage realistic. And in human medicine, real-world data networks such as TriNetX and Flatiron proved the model completely, building exactly this layer for human trials. No veterinary equivalent exists.
Why trial recruitment first
Trial recruitment is the wedge, not the destination. It is the use case where the pain is sharpest and the buyer is clearest, and where a sponsor will pay for an answer today. It also happens to be the use case that funds the hardest part of the business, which is assembling and structuring the data itself. Every practice connected for recruitment makes the dataset more useful for post-marketing surveillance, for prescribing and outcomes intelligence, and for the benchmarking that clinics themselves will pay for. The order matters: the long-term platform is only reachable through a first product someone wants now.
Built, not just described
The riskiest assumption in the whole idea is not commercial. It is whether messy, inconsistent, multi-language veterinary records can genuinely be turned into a searchable cohort without a human templating every format. So rather than describe that in a deck, it was built first. The feasibility engine now runs end to end on 320 consultation notes drawn from four different practice systems in two languages, grades its own accuracy against ground truth, and returns a cohort in seconds. That is the case example, and it is open for anyone to use.