Beyond Perplexity
Who buys structured, physician-authored clinical reasoning data? 14 prospects across health AI, frontier labs, and data infrastructure - grounded in links from Stealth and Lewis group chats, the Drive knowledge base, and Codex review.
1
Direct BuyersCompanies that write checks for clinical data
Each builds a product that generates or uses clinical reasoning. They need ground truth from real physicians.
Health Answer Engine
Perplexity Health
Consumer-facing health answers from clinical documentation. Need structured reasoning examples to evaluate and improve output quality.
Angle: 5-example sample package (patient profile + question chains). Bump this week.
In progress
Clinical Decision Support · $12B
OpenEvidence
AI consultation platform used daily by 40% of US physicians. 18M clinical consultations/month. Licensed data from NEJM, JAMA, AMA, Cochrane. $700M+ raised, $100M ARR.
Angle: They license journal content but lack physician-authored reasoning traces for eval. Coda fills that gap.
Healthcare Safety Agents · $3.5B
Hippocratic AI
Patient-facing AI voice agents for chronic care, follow-up, medication reminders. 7,500+ licensed clinicians have done 725K+ test calls. 50+ health systems across 6 countries. $404M raised.
Angle: They pay clinicians for safety eval already. Coda's structured traces scale this with provenance.
AI Medical Scribe · $5.3B
Abridge
AI clinical documentation from conversations. Mayo Clinic (2,000+ physicians), Duke, Johns Hopkins, Kaiser, UPMC, MSK. KLAS #1 in Ambient AI. Founded by UPMC cardiologist Dr. Shiv Rao.
Angle: Gold-standard physician-authored notes to benchmark AI-generated documentation quality.
Rx-to-Treatment Network · $1B
Forus
AI-powered prescription-to-treatment pipeline. Prior auth approvals in 2-3 days vs. months. $160M raised at $1B valuation. Founded by Sahir Jaggi (ex-Oscar Health). Free to doctors, monetized via biopharma/payer side.
Angle: Prior auth and appeals workflows need clinical reasoning traces with physician judgment.
2
Frontier LabsModel post-training buyers
Healthcare is the next domain where post-training matters. These labs already invest in clinical reasoning data.
Claude for Healthcare
Anthropic
Launched Jan 2026 at JPM. HIPAA BAA available. Connectors to CMS, ICD-10, PubMed, ClinicalTrials.gov. Customers: Banner Health (33 hospitals), Novo Nordisk, AstraZeneca. $380B valuation.
Coda fit: No disclosed clinical training data partnerships. Two-doctor model produces preference pairs they can't source internally.
HealthBench · ChatGPT Health
OpenAI
HealthBench: 262 physicians wrote 48,562 rubric criteria across 5,000 conversations. Formal healthcare division with 3 products. Customers: Cedars-Sinai, MSK, Stanford Medicine, HCA. 40M daily health queries.
Coda fit: HealthBench's physician-rubric format directly validates Coda's data shape. Strongest proof point.
Format validated
MedGemma · Google Health
Google DeepMind
MedGemma (open-weight, May 2025): 89.8% MedQA. Dedicated Google Health unit. Partners: Mayo Clinic, Stanford Medicine, Northwestern, CVS. MedLM deprecated Sept 2025; MedGemma is the active strategy.
Coda fit: Proprietary clinical training data partners not disclosed. Clinical preference data fills a gap.
3
Channels, Partners & Narrow TargetsDistribution, credibility, and niche workflows
Revenue through data partnerships, visibility through benchmarks, and specific clinical workflow buyers.
~$1.4B Revenue · Bootstrapped
Surge AI
SFT/RLHF/verifiers for OpenAI, Anthropic, Google, Meta. ~130 FTE + 50K annotators. No dedicated healthcare product yet. Edwin Chen (TIME100 AI 2025).
Revenue channel
HFC Medical Fellowship
Scale AI
Runs Human Frontier Collective Medical Fellowship sourcing board-certified MDs for clinical scenario design and model eval. $300M+ DoD contracts. Qatar healthcare AI deal.
Revenue channel
Healthcare AI Arena
Medical Sphere
Blind model comparison arena. Publish physician-authored cases as benchmark content for credibility.
Credibility
Eval Infrastructure
HuggingFace
Community Eval, leaderboards. Publish clinical reasoning eval benchmark for free distribution.
Distribution
Value-Based Care
Pair Team
CMS ACCESS participant. Chronic care triage eval data. Narrow budget, specific pitch.
Narrow workflow
Clinical Trial Matching
PACT AI
AI-driven trial recruitment. Trial eligibility eval data. Narrow budget, narrow pitch.
Narrow workflow
Market Thesis
1
Sarah Guo · Conviction
"Domain-focused AI companies will post-train. Coding was just first." Healthcare is the next domain where post-training matters.
2
Edwin Chen · Surge AI
Quality requires taste, depth, and expert judgment. The frontier labs that win understand what "good" means in their domain. In healthcare, that means real physicians.
3
OpenAI · HealthBench
Already uses physician-created rubrics and multi-turn health conversations. Validates that frontier labs buy exactly the data format Coda Health produces.
The sell
"We have the physician workflow that produces the structured clinical reasoning data you need to train, evaluate, and trust your health AI. Repeatably, with real doctors."
First outbound wedge
Private physician-authored multi-turn clinical reasoning eval pack
Target: OpenEvidence, Hippocratic, Abridge, Anthropic, OpenAI, Google
Proof: publish benchmark on Medical Sphere + HuggingFace
hello@codahealth.ai
Generated 05.13.2026 · Sources: iMessage, Google Drive, Lab repo, HuggingFace Evals Workshop, Surge AI interview, Codex review