On May 27, 2026, oncology AI startup Triomics announced a $22 million Series B led by Battery Ventures, with returning investors Nexus Venture Partners, Lightspeed, and Y Combinator all re-upping. The round lands two years after a $15 million Series A in May 2024, when the company was still a single-product bet on clinical trial matching. The company does one specific thing: use models trained specifically for oncology to turn thousand-page cancer records into structured answers that sit inside existing clinical workflows.
The trend behind the check is worth naming. Frontier models keep getting stronger, but healthcare systems pay for vertical AI that works in their specialty, plugs into their tools, and can be audited. CEO Sarim Khan’s growth numbers make the case: enterprise customers up fourfold in a year, annual recurring revenue up tenfold.
Oncology Records Are Too Long for Generic AI
Khan’s observation is blunt: as cancer patients live longer, their records keep growing — “we have seen medical records [with] thousands of pages of information.” Triomics’ core engine, OncoLLM, reads the entire chart — progress notes, pathology, imaging, molecular data, faxes — rather than summarizing page by page. That distinction matters clinically: a biomarker result buried on page 400 changes what page 12 means, and page-by-page summarization loses exactly those cross-references. The platform ingests HL7, FHIR, CCDA, PDF, and TIFF, integrates natively with Epic, OncoEMR, and iKnowMed, and is listed on the Epic App Marketplace.
The company claims its oncology-trained models beat generic AI agents at these tasks, which is why institutions like Memorial Sloan Kettering and Yale Cancer Center deploy them.
OncoLLM and Three Product Lines
Triomics splits its offering along the three most labor-hungry workflows in a cancer center:
- PRISM handles clinical trial matching: every scheduled patient is checked daily against every active trial, producing a prioritized, cited worklist. An ASCO 2025 abstract with an academic medical center reported roughly a 30% lift in accruals, 95% accuracy, and 40% more matches
- Symphony does visit preparation: it assembles a pre-visit snapshot of disease status, treatment timeline, and biomarkers, and drafts notes inside the physician’s workflow. Customers report 80% less pre-charting time for new visits and 45% for follow-ups
- Harmony covers cancer registry and quality abstraction: it generates NAACCR, SEER, COC, and QOPI submissions with citations per field, at 96% accuracy, 75% faster abstraction, and 60% faster case finding per a preprint
Customers and Operating Scale
Beyond MSKCC and Yale Cancer Center, Triomics’ site lists deployments including Mount Sinai (an AI trial-matching platform live since January 2026) and the Regenstrief Institute (a real-world evaluation begun February 2026). The scale numbers: over 120,000 charts read monthly, more than 65,000 patients screened monthly, and 10-plus NCI-designated cancer center customers — the company counts 4 of the top 10 US cancer centers.
Those are operational metrics, not demo statistics. You only get to 120,000 charts a month by being wired into live scheduling and EHR workflows.
Not Another Medical Scribe
The nearest competitors are AI scribes like Abridge and Microsoft’s Nuance, which summarize doctor-patient conversations. Triomics starts somewhere else: it processes the mass of historical records that exist before the conversation, and its output is not a narrative note but structured, cited, auditable fields. Every answer traces back to a source line in the record — a hard requirement for registry submissions where one wrong field means resubmitting. The two categories will likely collide, as scribes reach for more of the chart and vertical platforms reach for the visit itself, but for now the wedge between them is clean.
A Textbook Vertical-AI Play
From a $15 million Series A in May 2024 to this $22 million round, Triomics has run the standard vertical-AI playbook: pick a specialty where records are long, messy, and compliance-heavy, then sell the model, the integrations, and the workflow together. The lesson for product teams is that in an era of commoditizing general models, differentiation is not “can it summarize” — it is data-pipeline depth (HL7, FHIR, faxes), EHR integration (Epic Marketplace listing), and verifiable output (per-field citations). That is what healthcare AI procurement will buy over the next year.
Sources
- Triomics nabs $22M to bring oncology-specific AI to cancer centers — TechCrunch
- Triomics
- Triomics raises $15M for cancer clinical trials AI — TechCrunch
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.