A deterministic readiness engine for regulated engineering — with AI that is structurally incapable of fabricating a compliance claim.
"Is the evidence complete, consistent, and sufficient enough for a human to make an accountable decision?" Regulated engineering teams answer this by hand today — and the manual version doesn't get faster as headcount stays flat and review cycles compress.
Regulated engineering teams are under real pressure to move faster — and under real, justified fear of what an ungoverned AI copilot does to a safety-critical evidence trail. Most AI-in-compliance products today pick a side: move fast and hope, or stay manual and safe. This platform is built specifically to not have to choose.
The readiness score and the AI's opinion are never allowed to be the same thing. The separation is enforced in code, not just explained in a policy doc.
A fixed, published formula over structured, human-entered evidence. No model in the loop. Same inputs, same result, every time.
One local model, schema-constrained, every claim cited to an evidence ID, provenance-hashed, and technically unable to change the score.
Every readiness score traces back through a fixed pipeline. AI reads the same evidence and can advise alongside it — but never writes back into the chain.
IEC 62304, ISO 14971, IEC 60601, 21 CFR Part 803, and FDA design-control guidance modeled as reference data, not marketing copy.
No external API dependency in the core workflow. The AI integration validates at startup that its endpoint is local-only.
Four scenarios across three evidence-provenance categories — two public FDA recalls, a synthetic composite, and a synthetic regulatory reference — each returns a distinct, non-uniform score (82%, 70%, 74%, 69%), not one tuned demo number.
Architecture decision records, a release-review checklist, versioned release manifests, and 419 passing automated tests behind every change.
Four scenarios, three evidence-provenance categories, one unmodified algorithm. No case-specific scoring logic exists — the fact that every case lands on a different number is itself the proof.
Class I software-related recall (FDA record 212817). Stress-tests coverage and consistency scoring against real regulatory evidence.
Specification-conformance recall (FDA record 220257). A second independent case, confirming the pipeline generalizes.
The platform's regression anchor — reproducible, used in every automated test.
Synthetic case built around FDA design-control guidance, independent of any real product — tests evidence-completeness detection on its own.
Cannot create evidence, alter results, approve records, or declare compliance — by design, not by prompt.
Provider failures reject execution; they never silently fall back to a guess.
The nearest direct comparison — and the question any investor who knows this space will ask before we get the chance to explain it ourselves.
AI agents that automate validation, traceability, and FDA/EU MDR documentation directly. Already used by 3 of the world's top 5 medtech companies.
The readiness score is never AI-generated. AI cites, summarizes, and drafts — and is architecturally unable to write back into the score a reviewer relies on.
This sits inside the eQMS / design-controls category already occupied by Greenlight Guru, MasterControl, Arena, and Qualio — buyers who already budget for exactly this kind of tooling.
Up 35% year-over-year, the highest annual total since 2022. MedTech alone raised $3.7B in Q1 2025.
PTC acquired Arena Solutions for $715M (2021). MasterControl reached unicorn status on a $150M investment from Sixth Street Growth (2022).
No research firm independently sizes "design-controls software" on its own — sizing this exact wedge within the category above is a rigorous exercise still ahead, not asserted here.
Category comps price per-seat, sold annually. Per-seat pricing fits a QMS used by everyone, every day — it fits this product's actual usage pattern less well, so pricing is proposed per active design program instead.
A real premium for a real preference: regulated buyers already want their evidence to never leave the building — this platform is built that way by default.
The AI layer runs locally via Ollama, not a metered cloud API — marginal cost per customer doesn't climb with usage.
Sole builder of the platform to date: product architecture, the deterministic scoring engine, the governed AI gateway, and the regulatory domain modeling.
20+ years of R&D leadership in regulated medtech and aerospace, with a Ph.D. in Mechanical Engineering (Biomedical Optics). Most recently Senior Principal Engineer at Edwards Lifesciences, running the exact evidence workflow this platform organizes — IQ/OQ/PQ, CAPA, DFMEA, and ISO 14971 risk assessment under 21 CFR 820 — on $4M+ cardiac-device R&D programs. 27+ peer-reviewed publications, 9 patents, PMP.
Pre-seed, deliberately lean: the AI runs locally and stays strictly advisory, so there's no metered inference cost to fund and no case to make for AI holding decision authority — the ask covers a founder/first hire, two design-team pilots, and the infrastructure to support them, not a full go-to-market build-out. 2026 pre-seed rounds for AI startups typically run $500K–$2M, median around $1.2M; this sits at the lower-middle of that range. 12–18 months runway.
Proposed split, not decided — this is where your own knowledge of cost of living, hiring plans, and target runway should override the default above.
Full citations with exact supporting quotes: docs/marketing/Citations.md in the product repository. Summarized here for the room.
Ketryx $39M Series B (2025), $55M+ total — Ketryx press release; MedTech Dive; R&D World.
eQMS market size — Fortune Business Insights, Grand View Research. Digital-health/MedTech VC trends — HTD Health. Arena→PTC $715M — PTC's own SEC 10-Q filing. MasterControl $150M — MasterControl's own announcement.
MasterControl, Greenlight Guru, Qualio annual pricing — OpenRegulatory pricing analyses (three separate articles; none of the three vendors publishes a rate card).
2026 pre-seed round benchmarks — ValueAdd VC; AI-specific pre-seed guidance — Boilerplate Hub.
Slides 02–08 and 12–14 make claims about this platform's own architecture, current state, and roadmap — verifiable by reading the product repository (ADR-002, ADR-009, mvp/backend/ai_gateway/), not by citing a third party.