PAICON

PAICON is a health data and AI company focused on building globally representative medical datasets for use in research, diagnostics, clinical trials, and AI development. The company's stated mission centres on what it calls "the Remaining 84" — the approximately 84% of the world's patients whose health data is currently absent from most health AI training sets. PAICON develops infrastructure and platforms to ingest, harmonise, and make accessible multimodal health data from hospitals across more than 60 countries, with the goal of making AI-driven medicine work equitably across diverse populations.

The company serves pharmaceutical and biotech organisations, diagnostics and in vitro diagnostics (IVD) companies, academic medical centres and research institutions, and clinical settings. PAICON positions itself as a data-first AI company, building from the dataset layer upward rather than applying fairness corrections after the fact.

Platform Architecture: How PaiX Works

• Data Influx: Raw, multimodal data is collected from hospitals in whatever format each institution already produces, including scans, genomic reads, lab values, pathology slides, and clinical reports. No pre-filtering is applied, so no modality is excluded at the point of collection.
• PaiX Edge: An on-site data ingestion layer that sits inside the hospital. It connects to existing systems — scanners, labs, and records — and anonymises data before it leaves the institution. It standardises records on the way out and is designed to meet GDPR and local regulatory requirements by default, keeping the hospital in control of its own data.
• PaiX Atlas: A core data lake that aggregates public and proprietary data from 60+ countries into a single, comparable format. It links imaging, genomics, and lab data at the patient level, applies continuous AI-driven quality improvement, and grows monthly across disease areas.
• PaiX Navigator: The flagship platform where end users interact with the data. Users can filter by disease, population, geography, and modality; license validated datasets for AI training and research; query the system in natural language for prevalence, treatment response, or cohort fit; and run PAICON diagnostics within the same environment.

Data Coverage and Modalities

• Health data harmonised from more than 60 countries into a single foundation.
• Five modalities currently supported or in scope: pathology, radiology, clinical data (live), omics, and laboratory data.
• More than 130,000 public and proprietary cases queryable through PaiX Navigator.
• Data is standardised across scanner types, protocols, and languages, and linked per patient across imaging, genomics, and lab results.
• The dataset grows continuously, with AI agents applied to improve data quality over time.

Capabilities by Use Case

• Pharma and Biotech: Supports companion diagnostic (CDx) submissions, global trial design, and biomarker discovery. Specific capabilities include natural language querying of large case sets across 60+ countries, biomarker prevalence and treatment pattern analysis by geography, market sizing and competitive intelligence for business development, cohort sizing and stratification for trial design, and licensing of trial-ready cohorts with full data provenance.
• Diagnostics and IVD: Supports algorithm validation and benchmarking against diverse, geographically distributed datasets.
• Research and Academic Medical Centres: Provides access to multimodal, GDPR-compliant datasets for research purposes.
• Clinics and Hospitals: Offers patient-level insights and workflow integration tools.

Compliance and Deployment

• PaiX Edge is deployed on-site within hospital infrastructure, ensuring data anonymisation occurs before any data leaves the institution.
• The architecture is designed to meet GDPR requirements and local data protection regulations by default.
• Hospitals retain control of their own systems and data throughout the process.
• The platform connects to existing hospital systems without requiring institutions to change their formats or workflows.

Context and Rationale

• PAICON's founding premise is that current health AI has been trained on data representing roughly 16 out of every 100 patients globally, leaving the majority of the world's population underrepresented.
• The company's ByteSight content series features perspectives from researchers, clinicians, and data scientists on the consequences of this gap — including AI models that fail when applied to populations outside their training distribution, drugs whose efficacy or toxicity differs across ancestries, and genomic variants that cannot be interpreted for patients from underrepresented regions.
• Experts cited in the company's content include figures from Genome Canada, AstraZeneca, Queen Mary University of London, Genomics England, the University of Westminster, and Pixel Scientia Labs, among others, all speaking to the clinical and scientific risks of non-representative health data.
• PAICON was co-founded by Dr. Manasi A-Ratnaparkhe, who serves as CEO.

PAICON's approach involves building data infrastructure region by region and hospital by hospital, with the stated aim of expanding the representational base of health AI until it reflects the full diversity of the global patient population.