Patient data stays at the clinic.
Research insight travels.
Aggregate queries across distributed clinical sites — without centralising records, without moving PHI, and without a data engineering team.
Healthcare deployments typically involve custom workspace arrangements. Contact us before rolling out to multiple sites.
The problem
Multi-site research requires aggregation without centralisation.
PHI cannot leave the site
Centralising patient records — even temporarily — creates HIPAA exposure, IRB complications, and patient consent requirements that can derail a study before it starts.
Federated learning is complex
Proper federated ML infrastructure takes months to set up. Most research questions don't need a model — they need a row count, a distribution, or a cohort size.
Aggregate insight is the actual goal
Researchers need to know: how many patients match a cohort? What's the distribution of a lab value across sites? Which sites have missing data fields? Not the raw records.
How it works
Privacy by architecture — not just by policy
Each clinical site installs Sery Link
Site IT installs Sery Link on a designated analysis workstation. The workstation connects to local data exports (CSVs, Parquet files) and on-prem databases (Postgres, MySQL, ClickHouse, MongoDB) directly. Patient records never leave the site network.
Sites join a shared research workspace
The principal investigator's workspace connects all participating sites. Schema metadata — column names, row counts, value distributions — syncs to the coordination layer. PHI does not.
Query aggregate statistics across all sites
Ask "how many patients across all sites have a HbA1c value above 7.5?", "what is the distribution of age at diagnosis across sites?", or "which sites are missing the baseline_weight column?" — SQL executes locally at each site, only the aggregate result travels back.
Audit log satisfies IRB documentation
Every query and every outbound network event is logged locally at each site. The log can be included in IRB submissions as evidence that no patient-level data left the site.
What you'll do with Sery
The clinical drudgework AI was made for — on charts AI was never allowed near.
Find
Patients matching a clinical pattern across charts
"Show me patients over 60 with abnormal HbA1c trends in the last 90 days." Sery reads the charts on your machine and surfaces the cohort — no PHI export, no eDiscovery vendor.
Summarise
A morning stack of intake forms in minutes
Before clinic opens, ask Sery to summarise overnight intakes by chief complaint and risk flag. Read the briefing in three minutes; spend the rest with patients.
Draft
Referral letters and discharge instructions from chart notes
Sery reads your note in the EHR export and produces a first draft in your voice. You sign off; you don't restart from a blank page after every visit.
Compare
Lab trends across visits, without opening every chart
"How has Mr Henderson's creatinine moved across the last six visits, and what changed in his medication around each inflection?" One question, structured answer.
Reason
About cohort eligibility before writing a protocol
Sery checks every connected site's schema and patient counts against a draft inclusion list — and tells you which sites can actually contribute before you draft the IRB submission.
Aggregate queries
Five multi-site questions researchers ask
"How many patients across all sites match this cohort definition?"
Feasibility check before writing a full protocol
"What is the distribution of lab value X across sites?"
Site heterogeneity analysis without centralising records
"Which sites are missing the baseline_bmi column?"
Data quality check before statistical analysis
"Show enrollment numbers by site for the last 30 days"
Study progress monitoring across all participating sites
"Which sites have fewer than 50 patients with complete data?"
Power calculation and site-dropout risk assessment
Architecture
Patient data never leaves the site
Schema metadata only — no PHI in the coordination layer
Sery's cloud layer holds column names, row counts, and file paths. Patient identifiers, diagnoses, and lab values are never uploaded.
SQL executes locally at each site
When you query "patients with HbA1c > 7.5", the SQL runs on DuckDB at each site. Only the count (e.g., "42") travels back — never the patient list.
Per-site audit log for IRB documentation
Every outbound network event is logged locally. Site IT and IRB coordinators can review exactly what the system sent at any time.
AGPL-3.0 open source — verifiable claims
The desktop client is open source. Healthcare IT and compliance teams can audit the code, not just the marketing page.
Important: Sery is a coordination and query tool, not a certified HIPAA Business Associate by default. For regulated multi-site deployments, contact us to discuss a BAA and appropriate deployment configuration before going live.
Talk to us before you deploy
Healthcare deployments have compliance requirements that deserve a real conversation. We want to understand your use case before you connect patient data systems.
Sery Link is free to evaluate on non-PHI test data while you assess fit.