Capital for Cures runs a purpose-built AI research platform for pharmaceutical royalties. Owned hardware, no third-party cloud APIs, full data sovereignty. Three layers of compute and data designed for one job: surface every royalty-relevant signal in pharmaceutical filings, benchmark every disclosure against four decades of comparables, and ship the result as research the next morning. Daily ingestion. Every verified outcome feeds back into training.
Ingestion brings primary documents in. Analysis turns documents into a structured knowledge graph and runs the valuation work on top of it. Reporting surfaces the outputs as scored briefs, hidden-connection alerts, new-stream detection, and benchmark queries.
The system is partitioned by workload. Ingestion runs in parallel across primary sources. Enrichment runs domain-trained models against a forty-year deal corpus. A dedicated role handles non-English markets continuously. A central data layer serves the rest of the system and powers the scoring engine. We do not publish hardware specifics.
Parallel extraction · multi-source
Parallel extraction from SEC EDGAR, clinical trial registries (ClinicalTrials.gov, EU CTR), patent databases (USPTO, EPO, JPO), press releases, and annual reports. Every new filing is automatically parsed, classified, and queued for enrichment. Coverage runs daily across the source set.
Domain-trained models · multi-pass enrichment
Runs large parameter models purpose-trained on our deal corpus. Multi-pass enrichment: scientific context, competitive landscape, patent analysis, deal-precedent matching. Proprietary valuation engine runs DCF, Monte Carlo, and comparables against the full forty-year database for every asset.
Non-English filings · five-plus languages
Dedicated to non-English markets. EDINET filings (Japan), DART system (Korea), EMA regulatory data (EU). Custom Japanese NER and OCR pipeline for pharmaceutical deal extraction. Always-on coverage scanning filings in five-plus languages. This is how we find deals no one else sees.
Central data layer · scoring & screening
Central data layer serves all nodes. Deal scoring engine rates every asset on royalty and licensing attractiveness. Distressed-biotech detection (cash runway, management changes, declining market cap). Origination screening: which holders are most likely to monetise next. Surfaces results to the research team and the publication pipeline.
Royalty research touches private deal documents, draft term sheets, confidential disclosures from holders. None of that leaves our infrastructure. No third-party model provider trains on our corpus, and no inference call routes through anyone else's logging.
Our models are continuously fine-tuned on our own 40-year deal corpus. A general-purpose frontier model is excellent at language but does not know what a step-down clause looks like in a 1998 Merck-Schering term sheet. The domain training is the moat.
Daily ingestion across nine jurisdictions and five-plus languages would be uneconomic on per-token pricing. Owned hardware turns marginal cost to near-zero, which is what makes systematic coverage of sub-$100M positions tractable in the first place.
| Surface | What it is |
|---|---|
| Daily intelligence briefs | Scored and ranked against your investment thesis. Asset, action, why now. |
| Hidden connections | Entity relationships invisible in any single filing. Same molecule, multiple counterparties; same counterparty, multiple molecules. |
| New royalty streams | Buried in footnotes, litigation, amendments. Detected the day they appear in a filing. |
| Competitive landscape shifts | Pipeline, formulary, and pricing changes that move existing royalty economics. |
| Deal structure patterns | Rate benchmarks by therapy area, stage, structure. Where the cohort is tight; where it is wide. |
| Royalty rate calculator | Query fair rates by stage, indication, deal structure against the corpus. |
Numbers and architecture descriptions reflect the state of the engine as of the date of publication. The engine is a research tool. Outputs are research and analysis, not investment advice. Capital for Cures does not manage capital, hold positions, or place securities.
