The cancer slide already holds the answer. Ataraxis built a machine to read it - predicting recurrence and whether chemotherapy will actually help.
In a breast-cancer clinic, one of the hardest questions has nothing to do with whether a tumor exists. It is whether the patient in front of you will actually benefit from chemotherapy - or endure it for no gain. Ataraxis AI, a New York company founded in 2022, was built to answer that question with machine-read evidence rather than population averages.
The company describes its work as "AI-native" diagnostics. Instead of shipping tissue off for a molecular assay that consumes the sample, Ataraxis trains artificial-intelligence models to read the same high-resolution pathology slide a lab already produces. From that image it estimates the risk of cancer returning over five to ten years and, in its newest test, whether adjuvant chemotherapy is likely to help a specific patient.
The pitch is unusually concrete for an AI startup. Co-founder and chief executive Jan Witowski, a physician-scientist who trained at Harvard, Massachusetts General Hospital and NYU Langone, says a validation study found the technology roughly 30 percent more accurate than the current standard of care for breast cancer. Results, the company says, come back in about an hour and require no additional procedure.
That combination - higher reported accuracy, faster turnaround, and no extra tissue - is the wedge Ataraxis is using to enter a diagnostics market long dominated by genomic recurrence tests.
"For too long, chemotherapy decisions in breast cancer have been based on how a treatment works on average, not whether it will actually benefit the specific person sitting across from you."
- Jan Witowski, MD PhD, CEO & Co-FounderAtaraxis's stack starts with a foundation model and layers clinically focused tests on top. Each reads standard slides; each aims to change a treatment decision.
A digital-pathology foundation model pretrained on hundreds of millions of tissue images using vision transformers and self-supervised learning, with proprietary methods to scale and optimize it.
The prognostic and predictive platform for breast cancer. Combines imaging signals with clinical variables to assess recurrence risk, chemotherapy benefit, and neoadjuvant response.
The first predictive AI test estimating individualized chemotherapy benefit - modeling outcomes with and without chemo, powered by the Tau causal-inference layer that corrects for real-world bias.
Illustrative visualization of Ataraxis's public claim, not a peer-reviewed figure. Exact metrics depend on the study, subtype and endpoint.
Ataraxis sells into oncology. Its customers are the oncologists, surgeons and pathologists who order recurrence and treatment-benefit tests, along with the hospitals and cancer centers they work in. Since the 2026 launch of Breast CTX, the company says NCI-designated cancer centers and community clinics across the United States have begun adopting it. The business model is B2B clinical diagnostics: Ataraxis charges to return a result on tissue the clinic already holds, and works with trial networks such as France's Unicancer and the research organization MEDSIR to validate performance across diverse patient groups.
The problem it targets is overtreatment. In several breast-cancer settings, a meaningful share of patients receive chemotherapy that may not improve their outcome. Existing genomic assays help stratify risk but consume tissue, take longer, and leave gaps - notably subtypes with no recommended diagnostic tool at all. Ataraxis positions its image-first approach as faster, tissue-sparing, and able to work where those tools do not.
It is not alone. The company operates in a crowded field of AI-in-oncology and computational-pathology startups - names like Valar Labs and Manas AI have raised comparable sums - and against incumbent genomic-recurrence diagnostics. Its differentiation rests on owning a proprietary foundation model rather than fine-tuning someone else's, and on validating against, rather than alongside, the molecular standard.
Physician-scientist Jan Witowski and AI researcher Krzysztof Geras start Ataraxis AI.
The company launches Ataraxis Breast, backed by Giant Ventures and Obvious Ventures.
AIX Ventures leads, with Founders Fund, Thiel Bio, Floating Point and Bertelsmann participating.
The first predictive AI test for individualized chemotherapy benefit reaches cancer centers and clinics.
A physician-scientist with more than a decade in medical image processing, previously at Harvard Medical School, Mass General and NYU Langone. Frames the company as "an AI frontier lab, but for healthcare."
NYU Grossman assistant professor of radiology, affiliated with the Center for Data Science and the Courant Institute, with prior AI research at JP Morgan, Microsoft and Amazon.
Meta's chief AI scientist, pioneer of convolutional networks and a 2018 ACM Turing Award laureate, advises the company on artificial intelligence.
| Category | Precision-oncology AI diagnostics (B2B healthcare / SaaS-enabled) |
| Founded | 2022, New York, NY |
| Founders | Jan Witowski (CEO), Krzysztof Geras (CSO) |
| Headquarters | 169 Madison Ave, New York, NY 10016 |
| Team size | ~32 employees |
| Total funding | ~$24.4M ($4M seed + $20.4M Series A) |
| Lead investor | AIX Ventures (Series A) |
| Core tech | Kestrel foundation model; Ataraxis Breast; Breast CTX; Tau causal-inference layer |