Unison

Unison (legal entity Hyper Unison Ltd, London) operates a federated real-world evidence (RWE) platform. Rather than extracting and copying data into a central warehouse, it applies mappings at query time so that source systems retain their own schemas and vocabularies while researchers work against a single query surface. The company describes this as a 'virtual CDM' approach, contrasted on its site with batch ETL projects it characterises as taking 12-18 months per site and producing stale copies. A single query written in the company's query language (UQL) is translated into the SQL dialect of each participating site, executed locally, and returns aggregate counts, with no record-level data leaving the site [0][1].

Harmonisation is described in three layers: structure mapping (deterministic, schema-level rules written in SQL that map source tables and columns onto target CDM entities), semantic mapping (expert-curated resolution of source values to standard vocabulary concepts such as SNOMED or RxNorm), and value mapping (row-by-row logging so each source value can be traced back from an output number through concept, code and row, with the person, date and rationale recorded). Target models supported include OMOP as the default plus FHIR, CDISC SDTM, PCORnet and Sentinel. The site gives a worked example of 'Type 2 diabetes' expressed as ICD-10 E11.9 in US claims, Read CTV3 C10F. in UK primary care, SNOMED CT 44054006 in an EU hospital and a local registry code, all resolving to a single standard concept. Custom taxonomies, such as those needed for devices and diagnostics recorded as free text in hospital billing systems, can be ingested as first-class OMOP concepts [0][1].

An earlier positioning of the company, published by the PharmStars accelerator, describes Unison as a 'Genomics Data Brokerage Platform' letting private companies share genomic and microbiome data through a central platform that encodes, enriches and integrates omics data for pharma access, using a federated architecture that keeps raw data with the custodian while researchers train ML models and run analyses [2]. A company-registry summary similarly describes the business as AI data infrastructure for multi-omics to accelerate clinical trials, using genomic and proteomic data [3].