ClinicalContext — Join the Team

01 The Proof — What We've Actually Built

Not a prototype. Not a pitch. Measured, validated, patented mathematics.

0.614 Alpha (SNOMED CT) T(k) = A·k^α — the universal cost law

R²=0.988 Fit across 269 node pairs Published-quality result

38× Compression ratio <2% semantic loss

5 Domains validated SNOMED, NCI, BIM, Code + more

T(k) = A · k^α | α = 0.614 | R² = 0.988 The cost law governing medical knowledge graph traversal. Every clinical AI system is implicitly paying this cost. We're the only ones who've measured it — and compressed it.

What alpha = 0.614 actually means

Sub-linear scaling: Double the graph budget (k) and you get 2^0.614 = 1.53× more information — not 2×. The graph is compressible by design.
Not obvious, not trivial: If α were 1.0, no compression benefit exists. If α were 0, the graph is trivial. 0.614 is the Goldilocks zone — meaningful structure, exploitable compression.
Transferable property: We validated this across NCI Thesaurus (α=0.61), BIM ontologies (α=0.63), and code dependency graphs. The law holds. The patent covers the method.
38× at <2% loss: We don't just compress — we preserve semantic fidelity. The compressed subgraph contains the conceptually critical nodes. Sibling codes survive. Parent-child relationships survive.

The Patent: MVG-2026-001

Patent Title: Maximum Viable Graph — Budget-Constrained Subgraph Extraction
Filed: 22 February 2026
Reference: MVG-2026-001
Type: Method patent — the algorithm, not just an implementation
Coverage: α-calibrated traversal method applies to any medical ontology
PCT Extension: Planned — UK + US + EU
Scope: Competitor can clone the stack. Cannot clone the patent.

This is a method patent — it covers the mathematical approach, not just our specific implementation. A competitor building their own SNOMED compression tool using α-calibrated budget-constrained traversal is inside our patent boundary.

02 The Product — ClinicalContext API

One REST endpoint. 38× compressed SNOMED. Sub-100ms globally. Live now.

What it does

Takes free-text clinical phrase or partial code as input
Traverses SNOMED CT graph using MVG algorithm (α=0.614 calibrated)
Returns precise SNOMED CT code + confidence + hierarchy path
No SNOMED license required by the customer — managed server-side
X-API-Key auth, REST, JSON — drops into any stack in minutes

Infrastructure (live)

Deployed on Cloudflare Workers — global edge, zero cold-start
P99 latency: <100ms (sub-50ms typical at edge)
Gross margin: 97%+ — near-zero marginal cost per query
Auto-scaling: no DevOps, no capacity planning
Playground live on port 3005 — try it now

The Technical Architecture

Input: "shortness of breath with exertion"
↓ NLP extraction + intent parse
MVG: Budget-constrained subgraph extraction (k optimised, α=0.614)
↓ 38× compressed traversal of 350K SNOMED concepts
Output: SNOMED CT: 267036007 | Confidence: 0.94 | Path: [clinical finding → respiratory]
Time: <100ms end-to-end at global edge

Every ambient scribe processing a clinical encounter is implicitly doing this lookup — just badly. They use partial SNOMED tables, vector approximations, or fine-tuned LLMs that don't preserve ontological structure. We replace that with graph traversal that's provably correct and patent-protected.

Why existing approaches fail

Fine-tuned LLMs (Med-PaLM etc): No graph structure. Hallucination risk. Expensive at scale.
Vector embedding lookup: Loses ontological relationships. No hierarchy traversal. Sibling codes collapse.
Full SNOMED traversal: Computationally intractable at <100ms. 350K concepts × 1.5M relationships.
Build in-house: 12-24 months. $500K-2M engineering. No patent protection. Same structural blind spot.
ClinicalContext: 38× compressed. Patent-protected. Sub-100ms. 97% margin. Live.

03 The Market — Why This, Why Now

The ambient scribe market is exploding. Every one of those systems needs us.

$2.5B Ambient scribe market 2024 Nuance DAX, Abridge, Suki, Ambience

$12B Projected 2028 48% CAGR — fastest-growing clinical AI segment

10M+ Clinical encounters/day Each needing SNOMED precision

97%+ Gross margin Query pricing: $0.001–0.005

The customer is already known. The problem is already painful.

Ambient scribe vendors are not early-stage experiments. Nuance DAX (Microsoft) processes millions of clinical encounters daily. Abridge raised $150M. Suki raised $70M. These are real companies with real revenue and a real infrastructure gap.

The gap: Every ambient scribe needs SNOMED CT coding precision. None of them have a graph traversal layer. They're using pattern matching and hoping it's good enough.
The cost of the gap: Billing errors, CMS audit risk, EHR data quality failures. The liability is measurable.
Our position: ClinicalContext is the Stripe for clinical coding. One API call. Enterprise-grade medical reasoning underneath. Plug in once, never rebuild.
Expansion path: AutoCoder (RCM, $3.8B TAM) and PhenoMatch (rare disease) both use the same API. Platform leverage is real — we build the engine once.

Business Model

Layer 1 — Usage API: $0.001–0.005/query. Self-serve. 97%+ margin. Developer wedge.
Layer 2 — Enterprise License: $500K–5M/yr. Dedicated SLA, BAA, custom alpha calibration.
Layer 3 — Platform revenue: AutoCoder + PhenoMatch on same infrastructure. 2×–3× revenue multiplier.

Unit Economics at Scale

Nuance DAX alone: ~5M queries/day → $5K–25K/day at query pricing
1 enterprise license: $500K = equivalent of 500M queries. Enterprise is the real unit.
Year 1 realistic: 2 design partners + API usage = £150K ARR
Year 2 target: 1 enterprise contract + usage growth = £1.2M ARR

04 The Pitch Architecture — 15 Slides

The investor-grade story we've built. Expand each slide to see the content.

1 Vision — The infrastructure layer health AI is missing ▶

Every clinical AI application deserves full medical vocabulary precision — not vector approximations
ClinicalContext is the SNOMED CT API that ambient scribe vendors plug in once and never rebuild
Equivalent to Stripe for payments: one API call, enterprise-grade medical reasoning underneath
Target outcome: ClinicalContext powers >50% of ambient scribe queries globally by 2028

2 Problem — 10M clinical encounters/day, all with a SNOMED precision gap ▶

Ambient scribes (Nuance DAX, Abridge, Suki) process 10M+ clinical encounters/day — all need medical coding precision
Current approach: naive LLM inference + partial SNOMED lookup = sibling-coding errors, hallucinated codes
SNOMED CT has 350K+ concepts, 1.5M relationships — traversal at scale is computationally intractable without compression
Consequence: billing errors, compliance risk, downstream EHR data quality failures
Every ambient scribe vendor is building their own fragile SNOMED layer. Wasteful, slow, error-prone.

3 Insight — Medical knowledge graphs follow a universal cost law ▶

T(k) = A·k^α, α=0.614 for SNOMED CT — a power law governing graph traversal cost
The graph is compressible by 38× with <2% semantic loss — provably, mathematically
No other approach achieves this: vector embeddings lose structure; full traversal is too slow; fine-tuned LLMs don't generalise
Patent MVG-2026-001 protects the compression algorithm — not just an implementation, the mathematical method
Alpha transfers across medical ontologies (NCI confirmed): one patent, multiple addressable markets

4 Solution — Single REST endpoint, <100ms, 38× compression, live on edge ▶

Input: free-text clinical phrase or partial code; Output: precise SNOMED CT code + confidence + hierarchy path
Drop-in integration: no SNOMED license required by customer, no local graph infrastructure
Deployed: Cloudflare Workers (global edge, zero cold-start, sub-50ms P99)
Pricing: $0.001–0.005/query (B2C volume) + $500K–5M enterprise annual license (B2B SLA)

5 Market Opportunity — $2.5B → $12B, 48% CAGR ▶

Ambient scribe market: $2.5B (2024) → $12B (2028), 48% CAGR — fastest-growing clinical AI segment
Total addressable: every ambient scribe query = potential ClinicalContext call. Nuance DAX alone: ~5M queries/day
Serviceable (year 1): 3 ambient scribe vendors × 500K queries/day = $500/day → $180K ARR at entry price
Enterprise upside: 1 × $500K annual license = equivalent of 500M queries. Enterprise is the real unit economics.
Platform expansion: AutoCoder (RCM, $3.8B TAM) + PhenoMatch (rare disease, $1.2B TAM) use same API

6 Traction — Patent, API, R²=0.988, 5 domains validated ▶

Patent filed: MVG-2026-001 (22 Feb 2026) — method patent, not implementation
GraphContext API deployed on Cloudflare Workers edge infrastructure
SNOMED CT alpha validated: 0.614 (R²=0.988 across 269 pairs) — published-quality measurement
Transfer confirmed: alpha transfers across medical ontologies (BIM=0.976, Code=0.988, NCI validated)
5 domains validated — same MVP works for AutoCoder and PhenoMatch without rewrite

7 Business Model — Usage API + Enterprise License + Platform ▶

Layer 1 — Usage API: $0.001–0.005/query, 97%+ gross margin, self-serve onboarding
Layer 2 — Enterprise License: $500K–5M/yr, dedicated SLA, BAA, custom alpha calibration
Layer 3 — Platform revenue: AutoCoder + PhenoMatch built on same infrastructure (2×–3× revenue multiplier)
Unit economics at scale: 10M queries/day × $0.002 = $7.3M ARR (usage alone)
Blended target: 1–2 enterprise contracts + usage volume = $3–5M ARR by month 18

8 Competition — No direct equivalent exists. Moat is the patent + method. ▶

Direct: None with equivalent patent-protected compression. No comparable product exists.
Fine-tuned LLMs (Med-PaLM): no graph structure, hallucination risk, expensive to run
Vector embedding lookup: loses ontological relationships, no hierarchy traversal
Build-in-house: 12–24 months, $500K–2M in engineering time, no patent protection
Moat: patent (method, not implementation) + first-mover in compression primitive + alpha transferability

9 Go-to-Market — Abridge + Suki first, then enterprise, then platform ▶

Phase 1 (0–3 months): Design partner programme — 2 ambient scribe vendors (Abridge, Suki preferred — smaller, faster procurement)
Phase 2 (3–9 months): Usage-based self-serve launch, developer docs, SDK. Target: 10 API accounts, 1M queries/month
Phase 3 (9–18 months): First enterprise contract ($500K+). Target: one mid-market EHR or RCM company.
Channel: direct outreach to CTOs/VPs Engineering at ambient scribe vendors
Content: publish alpha validation white paper — positions ClinicalContext as the scientific standard

10 Technology / IP — MVG algorithm, Cloudflare Workers, SNOMED managed server-side ▶

Core: Maximum Viable Graph (MVG) algorithm — budget-constrained subgraph extraction with provable cost law
Patent: MVG-2026-001. Method patent covers the α-calibrated traversal algorithm. PCT extension planned.
Infrastructure: Cloudflare Workers (global edge, <100ms, auto-scaling, zero DevOps)
Validation: R²=0.988 (alpha), R²>0.94 (transfer at α-proximity <0.1) — publishable results
Defensibility: algorithm is the moat, not infrastructure. Competitor can clone stack, cannot clone patent.

11 Financial Logic — £1.2M seed, 48 months runway, break-even at 1 enterprise contract ▶

Seed raise: £1.2M / $1.5M
Burn rate: £25K/month (lean team, no office, cloud infra <£500/mo)
Runway: 48 months (to profitability or Series A)
Break-even: 1 enterprise contract ($500K) + 500M queries/year usage
Year 3 target: £5M ARR (3 enterprise + AutoCoder launch)

12 Use of Funds — Engineering 40%, BD 25%, IP 20%, Infra 10%, Reserve 5% ▶

Engineering (40% / £480K): SNOMED v2 (full 350K concepts), AutoCoder MVP, API hardening
Sales & BD (25% / £300K): First enterprise sales hire or fractional, conference presence, legal (HIPAA BAA)
Research & IP (20% / £240K): PCT patent extension, alpha validation papers, CIPA clinic follow-up
Infrastructure & ops (10% / £120K): Cloudflare scaling, SNOMED license, monitoring
Reserve (5% / £60K): contingency

13 Milestones — Month-by-month through Series A ▶

Month 1–3: 2 design partner LOIs signed. SNOMED v2 (full 350K) live. HPO alpha validated.
Month 3–6: First paying API customer. AutoCoder beta. White paper published.
Month 6–9: 10 API accounts. 1M queries/month. Enterprise pipeline of 5 qualified leads.
Month 9–12: First enterprise contract (£300K+). AutoCoder GA. Series A prep begins.
Month 12–18: £1.2M ARR. Series A raise (£5–8M target). PhenoMatch design partner.

14 Team — Rich (inventor), Steve (BD), and the gap we're filling ▶

Rich (CTO/CEO): Full-stack engineer. Invented MVG algorithm. Built and deployed GraphContext API. Filed patent. Domain: knowledge graphs, compression, clinical AI.
Steve Wing (CSO): Cambridge Law. Ex-JP Morgan VP. Strategic BD, legal, investor relations. Closes large deals.
Gap (flagged honestly): No clinical/health AI operator experience on team yet. Advisors being recruited: 1 practicing physician, 1 health AI operator.
Advisory target: Ex-Nuance or Abridge engineering VP — credibility with target customers.

15 Closing Thesis — One API call. Full clinical precision. The infrastructure layer that health AI is missing. ▶

The ambient scribe market will process 50M+ clinical queries/day by 2027. Every one needs medical coding precision.
ClinicalContext is the only patent-protected compression API that delivers full SNOMED CT precision at <100ms.
We have the algorithm (patent filed), the infrastructure (deployed), and the validation (R²=0.988).
£1.2M gets us to first enterprise contract and Series A readiness.
The alternative for every ambient scribe vendor: 18 months and $1M+ to build something worse.

05 The Team

Two people with complementary strengths. Built something real. Looking for one more.

Richard Woodman

Founder · CTO / CEO

Invented the MVG algorithm from first principles. Built GraphContext API solo — from concept to Cloudflare Workers deployment. Filed patent MVG-2026-001. Validated α=0.614 across 5 domains. Full-stack engineer with deep knowledge graph expertise.

The person who proved this works.

Steven Wing

Co-Founder · CSO

Cambridge Law. Ex-JP Morgan VP. Strategic BD, legal structuring, investor relations. Brings the institutional credibility and enterprise network to close deals that a pure-technical team can't. 50/50 equity. Complementary fit.

The person who can open doors at enterprise scale.

You?

Technical Co-Founder / Senior Engineer

We need someone who understands the math and can extend it. SNOMED v2 (350K concepts) needs building. The AutoCoder NLP pipeline needs designing. The white paper needs a co-author.

If you read the alpha derivation and had opinions — you're the person we're looking for.

Equity conversation, not salary conversation.

06 Where We're Going — 18-Month Roadmap

Concrete milestones. No fluff.

Month 1–3

2 design partner LOIs signed (Abridge, Suki preferred). SNOMED v2 full 350K concepts live. HPO alpha validated (PhenoMatch gate). White paper submitted to arXiv.

Month 3–6

First paying API customer. AutoCoder beta launched (ICD-10 accuracy gate: 90%). Developer docs + SDK published. 10 API accounts active.

Month 6–9

1M queries/month. Enterprise pipeline: 5 qualified leads. HLTH 2026 or ViVE 2026 conference presence. Clinical advisor signed.

Month 9–12

First enterprise contract (£300K+). AutoCoder GA. Series A prep begins. HIPAA BAA signed with enterprise customer.

Month 12–18

£1.2M ARR. Series A raise: £5–8M target. PhenoMatch design partner secured. PCT patent extension filed.

30-Day Critical Actions

Sign 1 design partner LOI — ambient scribe preferred (Abridge or Suki)
Publish SNOMED alpha white paper — credibility asset for all three products
Run HPO alpha experiment — 30-day binary gate for PhenoMatch
Identify clinical advisor — physician or CMO-level, shared across all products
Build investor deck — 15-slide from this architecture, ready for soft launch

Fundraising Timeline

Days 1–30: Pre-raise prep — LOI, white paper, 30 investor targets
Days 31–60: Soft launch — Steve leads warm intros (10 investors via JP Morgan/Cambridge network)
Days 61–90: Close — lead investor identified (£500K–800K anchor), fill with angels
Target: term sheet by day 90, round closed by month 4

07 Why Join Now

The window when joining a founding team is meaningful — and not risky because of nothing — is short.

This is the inflection point.

The algorithm is proven. The patent is filed. The API is live. The market is growing at 48% CAGR. What's missing is the team depth to execute on what's already been validated. That's a very different kind of early-stage risk than "we have an idea."

⚙️

The math is real

α=0.614, R²=0.988, 5 domains. This isn't a hypothesis. It's a measured property of medical knowledge graphs. We know it works.

📜

Patent protection

MVG-2026-001 filed 22 Feb 2026. Method patent. PCT extension planned. You're joining a company with IP, not just an idea.

⚡

API is live

Not a demo. Not a Figma prototype. A deployed Cloudflare Workers API you can hit today. The infrastructure exists.

📈

Market timing

Ambient scribes are at the S-curve inflection. Nuance DAX, Abridge, Suki — they exist, they're scaling, they have the problem we solve.

💰

£1.2M seed target

Raising seed in the next 90 days. Founding team equity is the conversation now — not employee options later.

🔬

Publishable science

The alpha validation methodology is JAMIA/arXiv calibre. Co-authoring the white paper is on the table. This is research that matters.

🏗️

Platform, not product

ClinicalContext is the base layer. AutoCoder and PhenoMatch are applications. One algorithm, three markets, one team that builds it.

🤝

Small team, real equity

Three people. Complementary skills. No bureaucracy. The decisions you make now will define the company. That's rare.

What we need from you

You can read T(k) = A·k^α and immediately think about what happens at the boundaries
You have opinions about knowledge graph architecture — and you'll push back when ours are wrong
You can build production-quality APIs, not just prototypes
You understand that SNOMED CT is not a flat dictionary — the hierarchy is the point
You want to co-own something that matters, not just get paid to ship tickets
You're comfortable with the fact that this is a seed-stage company — the upside is real because the risk is real

What we're offering

Founding team equity — meaningful stake, not employee options. Conversation based on contribution and fit.
Co-authorship on the alpha validation white paper (JAMIA + arXiv target)
Direct ownership of SNOMED v2 architecture, AutoCoder NLP pipeline, or PhenoMatch HPO layer — you choose what you build
Seed funding context: £1.2M raise in progress. Salary becomes the conversation once funded.
Cambridge Law + ex-JP Morgan + UK patent: the BD and legal foundation is there. You focus on the engineering.

Come build the infrastructure layer that clinical AI is missing.

The algorithm is proven. The patent is filed. The API is live. The market is ready. We're one person short of being able to execute on all of it.

Let's talk — ask us the hardest technical question you can think of

We built the missing infrastructurelayer for clinical AI.