Company · Life-Sciences AI · Y Combinator W26
A new drug can clear the lab and still wait weeks on a document. Ritivel's AI agents draft the regulatory submission - in Microsoft Word, formatted for the FDA, with every claim cited back to its source.
There is a version of drug development that never makes the news. A molecule works. The trial data comes back clean. And then, before any of it can reach a patient, someone sits down with thousands of pages of clinical results and starts to write. Not a press release - a submission. A document dense enough that regulators can decide whether a medicine is safe. That document can take weeks. Ritivel, a company in Y Combinator's Winter 2026 batch, is betting that AI can write the first draft.
The company calls itself an AI-native workspace for regulatory and medical writing teams at life-sciences companies. In plainer terms: it builds AI agents that produce the regulatory filings pharma companies have to submit - CTDs, CSRs, INDs, BLAs - and it produces them where the writers already work, inside Microsoft Word.
Ask most people what slows a new drug down and they will point at the trial, or the FDA, or the money. All real. But there is a quieter drag in the middle: turning a mountain of raw clinical data into the formatted, cross-referenced narrative that regulators actually read. It is expert work. It is slow. And every week of it has a price tag.
"Every month a drug is delayed from reaching the market costs pharma companies ~$45M in lost revenue."Ritivel
That number is the whole pitch, compressed. If a document is the thing standing between a finished trial and a shelf, then the document is worth speeding up - not by cutting the science, but by cutting the manual synthesis around it.
It helps to picture the actual task. A regulatory writer is not staring at a blank page waiting for inspiration. They are reconciling. A number in one table has to agree with a sentence three hundred pages away, which has to agree with a figure in a prior submission from two years ago, which has to be phrased the way a specific regulator expects. Multiply that across a full filing and the job becomes less like writing and more like auditing a small library by hand. It is precise, it is high-stakes, and it is the kind of work that quietly eats the most experienced people on a team.
Ritivel's agents generate a comprehensive first draft directly in Word, following FDA formatting rules and pulling from a company's existing clinical data and prior submissions. Three design choices make it more than a fancier autocomplete.
The one to underline is the middle-right box: citations. In a regulated field, a fast draft you cannot verify is worse than useless - it is a liability. Ritivel attaches a source link to each data point, which turns the human's job from "rewrite everything to be safe" into "check the ones that matter." That is the difference between a toy and a tool a compliance team will actually let near a filing.
The company frames the before-and-after bluntly: work that ran for weeks is meant to arrive as a draft in minutes. The point is not that the human disappears. It is that the human stops doing the copy-paste archaeology and starts doing the judgment.
A lot of enterprise software fails for a dull reason: it asks people to leave the tool they already trust. Regulatory teams live in Microsoft Word, store their evidence in SharePoint and Veeva, and chase approvals over Outlook. Ritivel's decision to draft inside Word, rather than in a shiny new interface, is a bet that adoption follows the path of least friction. The agent shows up where the work already happens instead of asking a compliance officer to learn a second home.
That choice also shapes what the product has to be good at. It is not enough to write fluent prose. The agent has to know which prior submission to borrow structure from, which trial table maps to which section, and how the FDA expects that section to look. Those are unglamorous, specific competencies - and they are exactly the ones a general chatbot does not have out of the box. The moat, if there is one, is built out of that specificity.
Ritivel is a business-to-business play. Its buyers are the regulatory affairs and medical writing teams inside pharmaceutical and biotech companies - the people whose calendars disappear into a submission for weeks at a time. The economic case writes itself against the company's own headline figure: if a month of delay costs roughly $45M, then software that shaves days off a filing does not need to be cheap to be obviously worth it.
As an early-stage company with three people, the customer base is design-partner scale, not a wall of logos. That is the honest shape of a Winter 2026 startup. The interesting question is not how many customers Ritivel has today, but whether the teams it does have will trust an agent with the one document they cannot afford to get wrong.
The three founders did not wander into pharma. Two of them - CEO Pavan Kalyan Tankala and CTO Nirmit Arora - came out of Microsoft Research, where they worked on language models and the safety of agentic systems. Tankala has published at NeurIPS, ACL and Interspeech and holds a BTech and MTech from IIT Bombay. The third, Gunin Gupta, handles business as CBO and came from consulting, also via IIT Bombay.
What is worth noticing is what they chose to build. People with that background usually ship a demo that gets a standing ovation on a conference stage. This team aimed the same firepower at regulatory filings - a problem the audience only understands if they have lived it. They got there the unglamorous way, through more than 50 conversations with pharma professionals, and kept following the complaint until it pointed at the submission.
The edge here is not the model. It is the willingness to sit in 50 conversations about a workflow no one brags about, and then build for it.
Regulatory information management is not empty territory. Veeva and other incumbents already run large parts of the life-sciences document stack, and general-purpose AI writing tools will keep drifting toward the enterprise. The default competitor, though, is the one that never shows up on a slide: the status quo of doing it by hand, in-house or through specialized vendors.
Ritivel's angle is to be narrow and deep rather than broad. Not a writing assistant that happens to work for pharma, but a workspace built around the exact documents, systems and formatting rules regulatory teams live inside. In a market that rewards trust and punishes shortcuts, running toward the regulated, high-stakes version of the problem may be the moat - because it is the version most AI startups would rather avoid.
There is also a timing argument. AI has gotten good enough at long-document reasoning that reading a thousand pages and holding them in mind is no longer science fiction, and regulated industries have started, cautiously, to let it in the door. The teams that were burned by early, over-eager AI tools are exactly the ones who will appreciate a product whose first instinct is to cite rather than to confidently invent. Ritivel is trying to be the version of this that a compliance officer can defend in a meeting.
The company is early. Three people, a Winter 2026 batch, seed money, a design-partner-scale customer base rather than a logo wall. But the thesis is legible, which is more than most. If the paperwork really is what keeps a working drug waiting, then the paperwork is worth automating - carefully, and with the receipts attached.