Oncology is a $250 billion US care market and the single hardest place to build production AI in healthcare.
The cancer centres that decide which infrastructure layer wins the next decade have already started picking.
Triomics, an oncology AI company helping cancer centers operationalize complex clinical information, has raised $22 million in Series B financing led by Battery Ventures, with participation from existing investors Nexus Venture Partners, Lightspeed, and Y Combinator, alongside strategic backers Oncology Ventures and Precision Health Informatics, a wholly-owned subsidiary of Texas Oncology. The round brings Triomics’ total funding raised to more than $36 million.
Triomics is already live at Memorial Sloan Kettering Cancer Center, MD Anderson Cancer Center, Yale Cancer Center and its partner Smilow Cancer Hospital, and Mount Sinai Tisch Cancer Center, alongside Texas Oncology, one of the nation’s largest community oncology practices. The customer list is the kind that takes years to build in healthcare and is itself the moat most enterprise AI startups have not been able to ship.
The Problem That Outgrew Legacy Healthcare Software
Cancer care is no longer constrained by missing information. It is constrained by the ability to use the information that already exists. A single patient history can span hundreds of pages of clinic notes, pathology and radiology reports, biomarker results, prior treatment summaries, outside records pulled in from referring institutions, and trial eligibility criteria that change month to month.
The workflows that turn this corpus into usable clinical context have stayed manual at most cancer centers. Research coordinators read records line by line to screen patients for trial eligibility. Registrars abstract chart data into state and federal cancer registries by hand. Oncologists prepare for visits by skimming charts they did not have time to fully read.
The information layer has scaled. The workflow layer has not.
The “show its work” part is what most healthcare AI has not solved. A summary that cannot be traced back to the source document is unusable in a clinical setting. A trial match that the research coordinator cannot verify against eligibility criteria does not get enrolled. Verifiability is the gating constraint, not raw model performance.
What the Metrics Say
The outcome numbers Triomics is bringing into the Series B are unusual for healthcare AI at this stage.
Published results show that users of the company’s product have increased trial matches by 40 percent and trial enrollments by more than 30 percent. Chart review times have come down by 67 percent. The platform has been peer-reviewed in Nature Digital Medicine and presented at the American Society of Clinical Oncology, the field’s premier clinical conference.
For context on why enrollment numbers matter at all: only a single-digit percentage of adult cancer patients in the United States enroll in clinical trials, a figure that has barely moved in two decades. The bottleneck has been identified repeatedly as the screening burden, not patient willingness. A 30 percent enrollment lift from a single software intervention, validated in peer review, is not a marginal improvement. It is the kind of number that converts AI adoption from a side project into a board-level priority for a cancer center.
What the Product Actually Ships
The Triomics platform is built around AI agents that read the full longitudinal patient record and convert unstructured information into structured, source-backed outputs. Three architectural choices distinguish the product from the summarization tools most cancer centers have already evaluated and rejected.
The first is the longitudinal record reader. Most healthcare AI reads a slice. Triomics reads the entire history, including outside records pulled in from referring institutions, biomarker results, and treatment notes spanning years.
The second is the verifiability layer. Every structured output the platform generates is tied back to the source document and the specific passage that justified it. The clinician reviewing a trial match or an abstracted data point can audit the reasoning directly inside the workflow.
The third is platform leverage. The same underlying AI infrastructure that powers proactive trial matching at MSK also powers pre-visit chart preparation, registry abstraction at Yale, and oncology data abstraction for quality improvement and operational use cases. One integration, multiple workflows.
Why Battery Led and Why Texas Oncology Joined
Battery Ventures has historically been one of the more selective enterprise software investors at Series B, with a portfolio that concentrates on infrastructure plays where workflow integration is the moat. The firm’s vertical AI thesis has been published consistently around infrastructure where the value lives in the workflow layer above the model, not at the model layer itself. Triomics fits the pattern.
The strategic investors are the more diagnostic signal. Precision Health Informatics, the wholly-owned subsidiary of Texas Oncology, is on the cap table as a strategic backer. Texas Oncology is also a customer. A customer-led strategic investment is a different signal from a financial investment. It implies the buyer expects the relationship to deepen across the next decade and is willing to put balance-sheet capital behind that thesis.
Oncology Ventures joining alongside is a sector-specialist endorsement from a firm whose pattern matching on cancer-care infrastructure is among the most informed in venture.
The signals to watch over the next twelve months: whether other large oncology networks follow Texas Oncology’s strategic investment model, whether Battery’s next vertical AI healthcare bet sits adjacent to Triomics or independent of it, and whether the company’s peer-review output cadence at venues like Nature Digital Medicine and ASCO continues at the pace it has set.
The Origin Story
Triomics was founded in 2021 by Sarim Khan, who serves as CEO, and Hrituraj Singh, who serves as CTO. The company spent four years building the foundation before reaching the scale of cancer-center deployment it announced this week.
The four-year build is the part that does not show up in the press release. Foundation models capable of reading thousands of pages of clinical narrative reliably did not exist in 2021. The product that ships in 2026 reflects a sequence of architectural decisions made when the underlying capability was years away from being commercially available.
Where the Bet Gets Decided
Three things determine whether Triomics converts from a high-performing oncology AI vendor into the default infrastructure layer cancer centers run on.
The first is whether the AI agent abstraction model replaces dedicated headcount, not just augments it. Registry abstraction is among the most labor-intensive workflows in a cancer center. The Yale quote describes a goal of “autonomous chart abstraction of clinical registry quality that can be rapidly reviewed and finalized for reporting by human registrars.” Whether the platform reaches the accuracy threshold where registrars review rather than create the output is the question that turns a productivity tool into a labor-model change.
The second is whether the life sciences pipeline materializes. The release describes a broader oncology network that helps life sciences organizations with their critical clinical-trial operations. Pharma and biotech customers buying into the same infrastructure that powers provider workflows would convert Triomics from a healthcare vendor into a two-sided platform sitting between sponsors and centers. The economics of that platform are materially different from the economics of a software-only business.
The third is whether peer-review continues. Nature Digital Medicine and ASCO are credentials that compound. Healthcare AI buyers discount vendor claims and over-weight peer-reviewed evidence. The platform’s ability to keep publishing at the pace the field demands is what determines whether the next ten cancer centers sign without lengthy custom pilots.
A platform that delivers all three would not be the largest healthcare AI company by revenue or the fastest growing by user count. It would be the company that proves the next phase of oncology care is won by who turns the patient record into shared, verifiable, actionable intelligence inside the workflow. The Series B is the bet that this is the company.
Don’t forget to like and share the story!
:::tip
HackerNoon has reviewed the report for quality, but the claims, opinions and analysis herein belong to the author.
All claims, opinions, and analytical conclusions are those of the author, who retains full control over the work.
#DYOR
:::