Healthcare technology startup BioStack has raised $9.5 million in seed funding to expand a clinical data infrastructure network spanning the United States, India, Brazil and Argentina. The company says its platform connects more than 100 million de-identified patient records, linking clinical notes, laboratory results, medical imaging, medications, treatments and outcomes. The goal is to give healthcare AI developers, biotechnology companies and researchers access to longitudinal clinical datasets for model training, evaluation and medical research.
BioStack Raises $9.5M to Build a Data Layer for Medical AI
BioStack is building healthcare data infrastructure designed to address one of the medical AI industry’s persistent challenges: fragmented patient information. The company announced a $9.5 million seed round alongside the launch of a network that aggregates and structures clinical information across hospital systems and international markets.
The funding round was led by Peak XV, Afore Capital, DCVC, Emerson Collective, SV Angel, Verdict Capital, Streamlined Ventures and Y Combinator. BioStack plans to use the capital to expand in the United States and other markets, develop its technology platform and grow its engineering and research teams.
The company’s approach focuses on connecting records that typically remain distributed across providers, clinical systems and geographic boundaries. These records can include physician notes, laboratory results, imaging, medications, treatment histories and subsequent patient outcomes. Linking those events over time can help researchers study disease progression and how patients respond to different interventions.
Connecting Fragmented Clinical Records
Healthcare data is frequently stored in systems that were designed for individual institutions or specific clinical functions. Even when information is available digitally, differences in formats, workflows and data access policies can make it difficult to assemble a consistent longitudinal view of a patient’s medical history.
BioStack says it develops workflows tailored to individual healthcare providers and their protocols, while creating proprietary datasets from de-identified information collected through those systems. The company describes its network as covering more than 100 million de-identified patient records across the four countries where it operates.
The distinction between record volume and usable data is important for AI development. A large dataset does not automatically provide reliable training material. Researchers also need consistent clinical labels, accurate links between events, appropriate representation of patient populations and sufficient context to interpret treatments and outcomes.
BioStack’s emphasis on longitudinal information could support research that requires more than isolated scans or individual clinical encounters. For example, a model intended to estimate treatment response may need information about a patient’s condition before treatment, the intervention received and subsequent clinical outcomes.
The company has not disclosed detailed independent validation results, dataset composition by disease area or the methodology used to measure data quality across its network.
Supporting Medical AI Training and Drug Research
BioStack says its infrastructure is designed to support medical AI model development and evaluation, clinical research, disease modelling, biomarker discovery, patient stratification and treatment-response analysis.
These applications are relevant as healthcare organisations explore machine learning for imaging analysis, clinical decision support and research workflows. Access to appropriately governed datasets can help developers train models and assess whether their outputs remain reliable across different patient populations and healthcare settings.
The company also reports that it has combined several hundred thousand X-rays, MRI scans, CT scans and related pathology and clinical records. According to BioStack, this multimodal collection could support the development of AI agents intended to help clinicians evaluate treatment options and potential patient outcomes.
That is a proposed use case rather than evidence of demonstrated clinical effectiveness. AI-generated recommendations require rigorous validation, and medical models must be assessed for bias, safety, reliability and performance in the populations for which they are intended.
BioStack’s focus on de-identified, longitudinal data may help address some research constraints, but data access alone cannot establish that a model improves patient care or produces better treatment decisions.
Privacy and Governance Remain Central
Clinical data infrastructure operates under stricter requirements than many general-purpose AI datasets. De-identification can reduce the risk of directly identifying individuals, but it does not eliminate every privacy risk, particularly when datasets contain detailed medical histories, dates, imaging or information that can be linked with external sources.
BioStack says its workflows are designed around patient privacy and individual provider protocols. The company has not detailed the specific technical controls, consent arrangements, retention policies or regulatory frameworks governing each national deployment in the announcement.
Healthcare providers and AI developers evaluating such a platform will need to understand how data is authorised for use, whether secondary research purposes are permitted, how cross-border transfers are managed and what safeguards apply to model training and evaluation.
Those considerations are particularly important for a network operating across multiple countries, each with its own healthcare data and privacy requirements.
India Programme Connects Clinical Data With Investigational Treatment
BioStack also highlighted work in India connecting cancer patients with an investigational cancer vaccine programme. The company says clinical information can support biotechnology development while eligible patients may gain access to personalised investigational treatments.
The programme illustrates a potential link between clinical data infrastructure and research recruitment. Properly governed clinical information can help researchers identify potentially eligible participants and understand disease characteristics relevant to a study. However, access to an investigational treatment does not establish its safety or efficacy, and eligibility remains subject to the applicable study criteria and clinical oversight.
Investors Back Healthcare Data Infrastructure
Peak XV Vice President Pushpak Kedia said access to high-quality longitudinal data will be important for the next generation of healthcare AI. The investment reflects interest in infrastructure that connects clinical records and makes them more usable for research and model development.
BioStack’s opportunity will depend on whether it can maintain reliable data pipelines, secure provider relationships and demonstrate that its datasets meet the requirements of researchers and AI developers. International scale may broaden the populations represented in its network, but differences in clinical practice, record quality and patient demographics also make harmonisation more complex.
The $9.5 million seed round gives the company capital to develop that infrastructure. Its longer-term prospects will depend on data quality, privacy governance, commercial adoption and evidence that the resulting datasets improve medical research and AI evaluation.
Market Landscape
Healthcare AI infrastructure is increasingly focused on data quality, interoperability and longitudinal patient information. Developers need datasets that connect clinical events over time, rather than collections of isolated records, to evaluate disease progression, treatment response and model performance.
BioStack’s approach sits at the intersection of healthcare data infrastructure, multimodal machine learning, clinical research technology and AI development platforms. Its international provider network could offer access to varied patient populations, although differences in clinical documentation, data standards and healthcare systems create significant integration challenges.
The regulatory environment is also becoming more focused on the quality and lifecycle management of medical AI. The U.S. Food and Drug Administration and its international regulatory partners published updated Good Machine Learning Practice principles in January 2025, emphasising the development of safe, effective and high-quality AI-enabled medical devices. FDA guidance
For BioStack, competitive differentiation will depend on the quality and representativeness of its datasets, privacy safeguards, provider relationships and the ability to demonstrate measurable value for research organisations and AI developers. The company’s reported 100 million patient records represent a substantial stated data footprint, but record counts alone do not establish clinical usefulness or model accuracy.
Top Insights
- Funding and expansion: BioStack secured $9.5 million in seed funding to expand its clinical data infrastructure, develop its platform and grow engineering and research teams.
- Multimodal medical AI: The platform connects clinical notes, imaging, laboratory results, medications and outcomes to support medical AI training and evaluation.
- International data network: BioStack reports access to more than 100 million de-identified patient records across the United States, India, Brazil and Argentina.
- Research applications: Potential uses include drug development, biomarker discovery, disease modelling, patient stratification and treatment-response analysis.
- Privacy and validation: De-identification and data aggregation do not eliminate governance risks; dataset quality, regulatory compliance and independent model validation remain critical.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
