Our client sells intelligence to pharma companies, researchers and investment analysts. Their raw material is research and trial data. Our team joined theirs to build and scale the platform that turns it into decisions.
The challenge
The data lived across public drug databases, clinical-trial registries, regulators and medical taxonomies. Much of it was free text: drug labels, trial protocols, eligibility criteria, regulatory documents.
Users needed it linked, current and searchable from whichever angle they start. Pick a disease, a drug, a molecular target or a technology, and see everything connected to it.
The data lake
A Ruby on Rails platform with screens for manual curation alongside automated workflows. A custom workflow engine runs each step in isolation on AWS Batch and Fargate.
Pipelines in Ruby, NodeJS and Python pull in and match drugs, trials and regulatory submissions. They sync trials often, keep drug regulatory data current, parse drug labels for indications, and add revenue and other figures to company profiles.

AI in the pipeline
LLMs turn free text into structured data at seven steps. They pull out entities and relations, structure indications and trial interventions (drug, dose, route), and turn eligibility criteria into fields you can query.
A further LLM step fact-checks the earlier extractions. Others link publications to trials, and read pharma news for PDUFA dates and licensing deals.
- Data scattered across sources→One linked, current dataset
- Free text nobody can query→Structured, fact-checked records
- One view per source→Four views: disease, drug, target, technology
The web app
Start from a disease such as breast cancer and see its statistics, trials and drugs in development. Start from a drug and see its stage, trials and competitors that share its targets. Search across everything with detailed filters.
A forecasting module models and compares market scenarios from disease incidence and trial progress, with saved scenarios and Monte Carlo forecasts.

Results
The platform now anchors strategic decisions for pharma companies, researchers and investment analysts.
- One dataset with four views: disease, drug, target and technology.
- A full view of drug development, from molecule to market.
- Free-text labels, protocols and criteria turned into fact-checked, queryable data.
- Market scenarios modelled from disease incidence and trial progress, with Monte Carlo forecasts.
- Records kept current by automated syncing and enrichment.