BioStack collects $9.5M to turn hospital records into AI training data
Hospital records are the raw material of medical AI and the hardest thing in healthcare to assemble at scale. BioStack just raised $9.5M in seed funding to try.
The round was led by Peak XV, with Afore Capital, DCVC, Emerson Collective, SV Angel, Verdict Capital, Streamlined Ventures and Y Combinator participating. The money goes toward US and international expansion, platform development and engineering hiring.
BioStack’s pitch is a data layer that follows a patient across systems. Notes, laboratory results, imaging, medications, procedures and outcomes get stitched into one longitudinal record, a view hospitals rarely hold because care fragments across providers and software vendors.
From that, the company builds proprietary de-identified datasets intended for three buyers. Clinicians get a more complete picture. Researchers get treatment histories tied to long-term outcomes. AI developers get multimodal training and evaluation sets that reflect real care rather than curated benchmarks.
Chief executive Sanat Mishra argues that capturing every event linked to a patient is what makes the stage-by-stage journey legible. Chief technology officer Parth Patwa leads the technical build.
The strategic bet is on a feedback loop. As more longitudinal records flow in, dataset quality improves, which attracts more research partners, which surfaces more data.
The risks are familiar. De-identification is legally delicate and technically imperfect, hospital data-sharing deals move slowly, and several well-funded competitors are chasing the same clinical corpus.
Still, the round signals investor appetite for infrastructure underneath healthcare AI rather than another model on top of it.
