Skip to main content
Platform / ProvenLake
Layer 1 — Core substrate

ProvenLake

Trust OS for Regulated Industries

Cryptographic chain of custody from raw data to regulatory filing. The same evidence chain satisfies FDA, TCFD/ISSB, NIST, and GxP — so the verticals built on top inherit defensibility by default.

In regulated work, the question isn't "what does the data say?" — it's "can you prove it?" ProvenLake answers that at the substrate level.

Evidence chain — substrate to application
Substrate
ProvenLake
.lake DSL compiler
Bronze→Silver→Gold
Cross-runtime determinism
Cryptographic attestation
Inherits defensibility
+
Phyto Drug discovery claims you can defend
+
UnitMesh Clinical decision units with FDA audit trails
+
Attesta SOC2 / PCI scan evidence chain
5
Domain Profiles
2
Runtime Engines
v0.1.0
Foundation Release
Prop.
Proprietary License

Platform Capabilities

Proprietary

ProvenLake is the trust layer that sits between domain applications and compliance reporting. Build the trust primitives once — evidence structure maps to every regulator you need.

+ Declarative pipeline definitions with the .lake DSL — domain profiles that extend vocabulary without touching the grammar
+ Dual-runtime execution (Rust for trust/attestation, Python for ML/enrichment) — same pipeline, same evidence, any runtime
+ Cryptographic chain of custody — deterministic evidence identifiers per step, pipeline attestation, append-only evidence store
+ Medallion architecture (Bronze/Silver/Gold) with lakehouse abstraction and backend plugins
+ Key lifecycle management — on-disk persistence with corruption detection, no silent rotation
+ Regulatory-body-agnostic trust primitives — evidence structure maps to FDA, TCFD/ISSB, NIST, and GxP disclosure requirements

Regulatory-Body-Agnostic Trust

The trust primitives don't know or care which regulator is checking. The same evidence structure maps to FDA, TCFD/ISSB, NIST, and GxP disclosure requirements.

Deterministic Evidence Identifiers

This exact computation happened with these exact inputs

FDATÜV SÜDTCFD/ISSB

Pipeline Attestation

Full execution chain signed by an authorized key

FDA (SaMD)CE (Annex II)NIST (SP 800-53)

Append-Only Evidence Store

No evidence was modified after the fact

FDA (21 CFR Part 11)ISO 13485SOC 2

Cross-Runtime Determinism

Same pipeline, same evidence — on Rust or Python

Any regulator requiring reproducibility

Key Lifecycle Management

Keys are persisted, protected, and corruption-detected

NIST (SP 800-57)ISO 27001

Domain Coverage

Domain profiles define vocabulary — kinds, relations, metrics — that pipelines operate on. New domains don't require compiler changes.

Profile Ready climate_infra profile

Climate & Insurance

TCFD & ISSB Evidence

Cryptographically attested climate risk pipelines. Every expected annual loss figure traces back to a verifiable pipeline execution — evidence structure designed for TCFD/ISSB disclosure requirements.

Pipeline: climate_infra_risk_engine
Simulation Validated clinical_operations profile

Clinical Operations

AI Governance for Healthcare

Trust layer used in UnitMesh's 50K-decision safety validation run with 0 adverse events. Clinical decision pipelines with immutable chain of custody. Targeting FDA De Novo clearance (DEN230156) by 2029.

Pipeline: LACE+ (discharge routing) · MARC (medication reconciliation) · RIRS (ICU readmission) · SEWS (sepsis early warning)
Profile Ready cyber profile

Cybersecurity

SOC 2 · PCI DSS · HIPAA · CMMC Evidence

Tamper-proof network scan evidence. Prove exactly what scan ran, when, with what inputs — auditor-ready for SOC 2, PCI DSS 4.0, HIPAA, and CMMC Level 2. NIST SP 800-53 and ISO 27001 coverage belongs to the BCM domain profile.

Pipeline: network_scan pipeline planned
Upcoming phyto (upstream) profile

Pharmaceutical R&D

FDA & GxP Computational Evidence

Computational evidence with the standing of a wet-lab notebook. Provenance columns trace every assay result to its pipeline step and execution run.

Pipeline: Bridges to 3it.dev.lake DuckDB
Commercial Licensee First commercial ProvenLake license · healthcare field of use · non-exclusive

Simulation Reference: UnitMesh

UnitMesh is a clinical decision simulation platform — not a 3IT product, but the application that exercised the full ProvenLake stack in production conditions. 50K decisions, 0 adverse events, 4 hospital scenarios. Targeting FDA De Novo clearance (DEN230156) by 2029.

50K Simulation Decisions 0 Adverse Events FDA De Novo DEN230156 · Target 2029
Four Decision Units — what ProvenLake attested
LACE+
Discharge Routing
Risk score · patient feature snapshot · clinician routing decision · timestamp
MARC
Medication Reconciliation
Drug interaction flags · reconciliation action · practitioner ID · timestamp
RIRS
ICU Readmission Risk
Risk prediction · contributing factors · care plan trigger · timestamp
SEWS
Sepsis Early Warning
Vital sign pattern · alert threshold crossed · clinical response time · timestamp
Full stack — where all three layers converge
RAD
Methodology
Rapid Application Development discipline — how units are specified, simulated, and submitted
Blue
Pipeline Runtime
Executes the decision pipeline — deterministic, auditable, runtime-agnostic
ProvenLake
Trust Layer
Cryptographically attests every decision record — evidence the FDA auditor can verify independently
UnitMesh is the only application currently running the full stack in production-equivalent conditions.
Visit unitmesh.dev →

Build on proven trust infrastructure

Proprietary platform. Start with the .lake DSL and DAG executor — add cryptographic attestation and regulatory export as you need them.

Request Demo