Biotech & drug discovery · Biomanufacturing software
AbInitio Bio
A two person Boston startup out of Y Combinator's Spring 2026 batch, building foundation models meant to predict how a biologic drug will behave in manufacturing before a single batch is run. No named pharma or CDMO customer has been identified yet.
Watching a drug die in the plant, not the lab
Daniel Mukasa spent years designing molecules, first at Caltech, then in a Merck internship, then in postdoctoral labs at MIT and the Broad Institute. The molecules kept clearing the science. Then, by his own account of starting the company, they kept failing somewhere else. “I’ve spent much of my career at Caltech, Merck, MIT, and the Broad doing drug discovery, just to watch our carefully optimized drugs get killed by manufacturing problems,” he wrote in AbInitio Bio’s Y Combinator launch post. “I’m building the tools we need to fix that across the full biomanufacturing stack.”
That is the pitch: not a better way to find a drug candidate, but a better way to predict, before a company commits a scale up campaign, whether that candidate can actually be made at volume.
The problem, in the company's own numbers
The launch post states: “A single CHO cell line campaign costs $5-10M and 12-18 months. Multispecific antibody programs routinely fail at CMC. Every month of delay on a blockbuster costs pharma $100M+ in lost revenue.” Pharma, it says, “spends over $100B annually on biologics development and manufacturing, and ships processes with yields and timelines that haven’t fundamentally shifted in a decade.”
The stated reason: every product team starts over, because “nothing transfers across products, scales, or sites,” and the process data that might change that “sits trapped in PDFs, batch records, and proprietary silos.” AI has poured into drug discovery for years; AbInitio Bio's framing is that manufacturing, the step deciding whether a discovered drug reaches patients, has been left behind.
From postdoc to R2E to YC
Mukasa's path from lab to founder runs through a specific MIT program. In an interview published by MIT HEALS, he described his title at MIT as “the MIT-Takeda Distinguished Postdoctoral Fellow working in MIT ChemE/MechE and the Broad Institute.” Through MIT's Researcher 2 Entrepreneur, or R2E, program, he began testing whether his machine learning work could become a company. “My R2E project initially focused on using machine learning for drug discovery,” he said. “Over the course of the program and through much advising, I broadened my horizons and tried to work backward from what the industry wanted to buy rather than what novel research had to offer. I think this was one of the most important steps for me in understanding what it takes to make a successful business.”
“Since doing R2E I’ve started my company Abinitio Bio, been accepted to the Y Combinator program, and engaged with people across pharma and biotech to understand their real-world problems,” he told MIT HEALS. A separate third party research profile, citing the company's own site, puts the R2E support at up to $20,000 and notes AbInitio Bio was also selected for Fifty Years' 5050 program in June 2025; Fifty Years describes 5050 as a free program with no funding or equity exchanged, so it is not a financing event.
Y Combinator accepted the company into its Spring 2026 batch, with Nicolas Dessaigne listed as primary partner. Break VC, the venture firm formerly Breakthrough Ventures, separately lists AbInitio Bio in its portfolio, describing Mukasa's record: a Caltech PhD in applied physics and materials science, a postdoctoral fellowship at MIT where he was “recognized as a Distinguished Fellow and finalist for the Alzheimer’s Insights AI prize,” and research “cited over 1,800 times with publications in Nature Materials, Nature Biomedical Engineering, Advanced Materials, and Chemical Reviews.” No source found states an investment amount.
Building the paperwork alongside the model
A customer story published by the legal technology vendor Fearn describes Mukasa filing a patent application for AbInitio Bio without the in-house counsel he had at Caltech and MIT. “When you leave these large institutions and their in-house legal counsel, you don’t have the same support anymore,” he told Fearn. “Filing a patent application can take weeks if not months now because you’re just another company in a Big Law firm portfolio. If you’re a fast growing startup like us, that timeline could kill you.” It is a small data point, but independent confirmation that the company is filing intellectual property this early, largely on its own.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| Named customers or pilots | None found. The launch post says its models can be adapted by “any pharma or CDMO,” and Break VC's page says the company “partners with biopharmaceutical companies and CDMOs,” but neither names a partner, and no pilot announcement or press release naming one was found as of Sep 24, 2026. | Not found |
| Regulatory and clinical records | No FDA 510(k), PMA, registered trial, NIH award, or PubMed listing under AbInitio Bio was found. | Not found |
| Federal grant records | A name based NSF award search returned only awards to unrelated companies with “Bio” in their name. No NSF, NIH or SEC record tied to AbInitio Bio was found. | Not found |
| MIT R2E grant | MIT's own HEALS profile confirms Mukasa received an R2E award that led directly to starting AbInitio Bio and joining Y Combinator, though it states no dollar figure. A third party profile puts the award at up to $20,000, citing the company's site. | Independent |
| Academic and professional background | A Caltech PhD, an MIT-Takeda Distinguished Postdoctoral Fellowship in MIT ChemE/MechE and the Broad Institute, and a prior Merck internship are confirmed by MIT's HEALS profile and, separately, by Break VC and Fearn's customer story. | Independent |
| Product description | AbInitio Bio's site and launch post describe a foundation model, Echo, predicting biomanufacturing outcomes across process development, scale up, manufacturing and quality. An earlier third party profile (Whiteford Research Biobase) describes a differently named product, “Therascript,” focused on generative antibody design. That profile carries its own disclaimer about possible AI generated inaccuracies, and its description was not confirmed elsewhere. | Differs from company site |
| Investors | Y Combinator (Spring 2026, primary partner Nicolas Dessaigne) and Break VC are named on their own sites as backers. No specific dollar amount and no SEC Form D was found for the company. | Amount not found |
Read plainly: the parts of the story that independent sources can check, mainly Mukasa's academic background, hold up. The parts that would show the product works in the field, a named customer, a published pilot result, an outside audit of Echo's predictions, do not yet exist in public sources. That is ordinary for a two person company a few months out of an accelerator, but it means this page rests heavily on what the company and its investors say about themselves.
What to watch
- A named pilot, CDMO partnership or paying customer, which nothing in this record identifies yet.
- Whether the company settles on one public description of its product, given the older, differently named description found in one third party profile.
- Growth beyond a two person team, Y Combinator's listed current size.
- An SEC Form D, which would be the first public confirmation of a specific investment amount.
In their words
“Biomanufacturing is bottlenecked by experiments that take years and don’t transfer across products. We’re building foundation models for biological manufacturing, validated against wet lab data, so decisions that take 6-18 months today can be made in hours, saving $100M+ per program.”
Daniel Mukasa, founder and CEO, Y Combinator launch post, 2026 · Company
“I’ve spent much of my career at Caltech, Merck, MIT, and the Broad doing drug discovery, just to watch our carefully optimized drugs get killed by manufacturing problems. I’m building the tools we need to fix that across the full biomanufacturing stack.”
Daniel Mukasa, Y Combinator launch post, 2026 · Company
“My R2E project initially focused on using machine learning for drug discovery. Over the course of the program and through much advising, I broadened my horizons and tried to work backward from what the industry wanted to buy rather than what novel research had to offer.”
Daniel Mukasa, to MIT HEALS, 2025 · Independent
“When you leave these large institutions and their in-house legal counsel, you don’t have the same support anymore. Filing a patent application can take weeks if not months now because you’re just another company in a Big Law firm portfolio. If you’re a fast growing startup like us, that timeline could kill you.”
Daniel Mukasa, to Fearn, 2026 · Independent
“Strong academic pedigree (Caltech PhD, MIT postdoc, Merck ML internship) but this is essentially a solo academic founder with no commercial product, no co-founder, no customers, and no visible traction.”
The YC Tier List, assessment of AbInitio Bio, 2026 · Independent
Related companies
Sources
- DirectoryAbInitio Bio company profile
- CompanyLaunch YC: AbInitio Bio, frontier foundation models for biomanufacturing
- IndependentHumans of HEALS: Daniel Mukasa, R2E Award Recipient
- InvestorAbInitio Bio portfolio page
- Vendor case studyAbInitio Bio is filing for priority on the science's timeline, not its law firm's
- IndependentAbInitio Bio assessment
- UnverifiedAbinitio Bio company profile, including Therascript product description and funding estimate
- CompanyAbInitio Bio website
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
