One domain done right. Then the next.
Building a verified evidence base is not generic infrastructure — each domain requires its own source set, ontologies, calibration methodology, and regulatory grounding. We build one domain to depth before expanding.
The arc
Medical
LiveClinical evidence, genomics, pharmacology, oncology, and rare disease — ingested from 143 authoritative sources, cross-validated, and calibrated. Nine products live.
- Clinical literature (PubMed, Cochrane, bioRxiv)
- Genomics & variants (ClinVar, gnomAD, GWAS Catalog)
- Drug & pharmacology (ChEMBL, DrugBank, CPIC)
- Oncology (TCGA, CIViC, COSMIC)
- Adverse events (FDA FAERS, OnSIDES, VAERS)
- Disease ontologies (MONDO, HPO, Orphanet)
- Skippy Ground
- Skippy Auth
- Skippy Variants
- Skippy Med-Check
- Skippy Rare
- Skippy Integrity
Legal
Active DevAlpha · Q3 2026 est.Case law, regulatory text, contract language, and compliance frameworks — verified against authoritative legal sources with citation to specific holdings and statutory sections. 31 ingest sources active.
- Case law: Harvard Caselaw Access, CourtListener, ECHR
- Regulatory text: CFR, Federal Register, EUR-Lex
- Statutes: USC, UK legislation, Canadian federal law
- International: UN treaties, WTO disputes, ICSID awards
- Contract: CUAD annotated dataset, FIBO ontology
- Contract verification
- Regulatory compliance intelligence
- Case law grounding
Scientific Research
Active DevAlpha · Q4 2026 est.Research literature, preprints, and experimental findings — cross-validated across sources with explicit uncertainty quantification for contested or early-stage findings. Ingest modules active.
- Preprints: arXiv (all categories), bioRxiv, medRxiv
- Physics literature: INSPIRE-HEP
- Academic graph: OpenAlex, Semantic Scholar
- Formal mathematics: Lean Mathlib, Metamath, OEIS
- Research datasets: GEO, ENCODE, PhysioNet
- Research claim verification
- Dataset grounding
- Literature synthesis
Finance
In SpecIn specification · 2027 est.SEC filings, earnings disclosures, macroeconomic data, and market information — verifiable financial claims with source lineage to original filings and regulatory documents. 6 ingest sources active.
- SEC EDGAR: 10-K, 10-Q, 8-K filings
- Macroeconomic: IMF, World Bank, Federal Reserve (FRED)
- Central banks: ECB Statistical Data Warehouse, BIS
- Market data: equity prices, indices, ETFs
- Financial standards: GAAP, IFRS codifications
- Financial claim verification
- Earnings intelligence
- Regulatory filing analysis
Working in Legal, Scientific Research, or Finance?
Enterprise teams with active deployment needs in upcoming domains help shape feature sequencing and get early access to alpha releases. If your compliance problem is in a domain we have not yet shipped, we want to hear from you — customer use cases directly influence what ships next.
Horizontal capabilities that unlock after the core domains ship
These products are not domain-specific — they apply Skippy's verified-belief infrastructure across the platform once Medical and Legal are fully operational. Each is in specification; none are in active development yet. Estimated timelines shown below are contingent on Medical and Legal reaching production stability.
Replace static rule tables that break every time a guideline changes. Every business rule — prior-auth criteria, underwriting exclusions, eligibility determinations — executes against a live, versioned evidence graph. When a CMS guideline updates, rules re-evaluate automatically against the new source. Adds first-class CONTESTED and NOT_COVERED outcomes to rule evaluation. Targets insurance underwriting, prior-auth, AML/KYC, government eligibility, and trade compliance.
Stop fabricated CVEs and hallucinated IOCs from entering your incident response chain. Verified-belief grounding for CVE reasoning, TTP attribution, and IOC validation — every finding traceable to NVD, MITRE ATT&CK, or vendor advisory with a signed audit record for post-incident regulatory reporting. Calibrated severity scores replace expert-hedged LLM outputs. Targets SOC teams, MSSPs, and CTI platforms.
Multi-policy, multi-jurisdiction Trust & Safety reasoning for VLOPs and UGC platforms. Produces DSA Article 17 Statements-of-Reasons with verifiable rationale — the audit artifact regulators require under the Digital Services Act. Handles CSAM, TVEC, NCII, deepfakes, and hate speech with per-jurisdiction policy lensing. Designed for 10K decisions/second at regulatory defensibility.
Graph-based per-user and per-organization memory layer for enterprise AI — with GDPR right-to-erasure and HIPAA data-segmentation built in. Retains verified beliefs about users (preferences, permissions, clinical history) in an auditable, versionable graph. Gives compliance teams a durable record of what the AI knew about a user at each decision point.
Verifier-aware translation that preserves citation chains, regulatory terminology, and audit-trail integrity across languages — so a German regulator receives the same evidence lineage as an FDA auditor. Wraps commodity translation engines with cross-lingual NLI to validate that translated claims still entail their sources. Per-language ECE calibration. Tier-1 launch: EN, DE, FR, ES, IT, NL, PT, JA, ZH-CN, KO, AR.
Why not everything at once?
Verification quality is domain-specific. The evidence hierarchy for a genomic variant claim — ACMG/AMP criteria, gnomAD frequencies, functional studies — is entirely different from a legal claim citing jurisdiction, precedent, and statutory text. Generic verification produces generic quality.
Each domain requires the right source set, the right ontologies, the right calibration methodology, and the right regulatory defensibility model. That work cannot be shortcut without compromising the verification contract.
Medical came first because it has the highest stakes — clinical AI errors cost lives. Every domain that follows inherits the same verifier architecture and audit layer. The depth comes first; the breadth follows.
- We will not ship a domain without a calibrated, held-out evidence base. Source count alone is not a launch criterion.
- We will not expand to a new domain before the preceding one reaches production stability — even if the market pressure to do so is high.
- We will not build platform extensions (Rules Engine, Memory, Localization) until the domain foundations that justify them are operationally proven.
The verifier, the audit layer, and the confidence model are shared infrastructure.
Every domain that gets added inherits the same verifier contract, the same immutable audit trail, and the same calibrated confidence model. A legal query and a medical query pass through the same verifier — the domain-specific difference is the source set and the calibration methodology. That's what makes this a platform, not a collection of products.
Working in one of these domains?
We are actively sequencing the domain roadmap. If you have a use case in Legal, Scientific Research, or Finance, we want to hear from you.
Get in Touch