AI research agents that work like an expert team — on your data, in your infrastructure. Federated execution means your data never leaves your walls.
Need data? EvidenceKit pairs the Analytics platform with all-payer claims data.
An agentic loop with human oversight at every step. Queries run on your data — and execute in your environment or ours, depending on your preference.
"What is the impact of Drug X on patients with Y?"
A sample of what researchers ask — and the answers they get back.
We benchmark the pipeline at every checkpoint — from how we interpret your question to how we verify the final numbers. No black boxes.
Every feature is validated against peer-reviewed research before it ships. We reproduce published results and compare outputs to ensure accuracy.
Every dataset is different. Onboarding benchmarks validate the pipeline against your specific data structure and coding practices before you run a single query.
Every query produces a complete audit trail — traceable, reviewable, reproducible. Every cohort definition and statistical choice is documented.
Every stage is benchmarked against gold-standard references, published literature, and expert-curated annotations.
Never. Medeloop uses a federated execution model — the compute goes to your data, not the other way around. Your data stays in your infrastructure, and only aggregated, de-identified results are returned. No raw patient data ever leaves your walls.
Yes. Medeloop deploys in your environment (AWS, Azure, GCP, or on-prem) or in the Medeloop cloud — you pick based on your compliance, security, and operational preferences. In either configuration, queries run on your data and only results are returned.
Medeloop works with EHRs, claims datasets, disease registries, labs, and institutional databases — including custom schemas. If you have structured clinical data, Medeloop can run on it.
No. Our semantic engine reads your data natively — no OMOP mapping, no CDM conversion, no months of prep. If you already have OMOP, I2B2, or another common data model in place, Medeloop works with those too.
No coding required to get started. You describe your research question in plain English and the agentic pipeline handles cohort definition, query execution, statistical analysis, and output generation. For technical teams, every step is fully inspectable — with editable code, full audit trails, and the ability to drop into Python for deeper customization.
Every stage of the pipeline is benchmarked against gold-standard references, published literature, and expert-curated annotations. We validate six checkpoints — query understanding, concept extraction, code mapping, cohort construction, statistical analysis, and result verification. A third-party validation paper is available on request.
Each study produces a defined patient cohort, the full query execution log, statistical outputs (Kaplan-Meier curves, regression tables, descriptive statistics), data visualizations, and a manuscript-ready narrative report. Every output is traceable, reviewable, and reproducible — designed to hold up to peer review, IRB review, or board presentation.
Most analyses complete in minutes. Complex multi-step studies with large cohorts may take longer depending on your data infrastructure. The agentic pipeline runs all steps automatically — you review the plan, approve it, and the system handles execution.
Analytics is the platform — it runs on your own data. EvidenceKit pairs the Analytics platform with the licensed HealthVerity all-payer claims dataset in a single subscription. If you don't have your own institutional data yet, EvidenceKit is the fastest way to start.
Real-world evidence at the speed of a question.
