Compute marketplace — federated training + inference
Lend your compute. Earn HCC for verifiable work.
The AI master model behind Conceptual Health (orb scoring, risk stratification, the user-AI assistant) is trained federally — compute contributions come from clinics, opt-in patients with capable hardware, and Conceptual Healthcare Corporation's own fleet. Every verifiable cycle earns HCC at a published per- workload rate. Every workload carries a privacy manifest that determines which contributor tier can run it.
Three contributor tiers
Patient home compute
Opt-in · CPU + 16GB RAM minimum
Modern desktop / laptop hardware (CPU + ≥ 16 GB RAM + optional GPU). Opt-in via the patient app. Energy-cost preference (don't mine on battery, don't mine above a kWh threshold). Eligible workloads: public + synthetic data training, federated training on your own data, FHE / MPC analytics, knowledge-class user-AI inference, public-benchmark verification, synthetic-data generation.
Clinic compute
TEE-attested · HIPAA-covered
On-premise clinic servers with Intel SGX / AMD SEV / Apple Secure Enclave remote attestation. Eligible for all workloads including cleartext PHI within the clinic's care relationship. Off-hours utilization typical; HCC earnings flow to the clinic's wallet.
Corporate compute
TEE-attested · Same rate as clinics
Conceptual Healthcare Corporation's own TEE-equipped fleet. Earns HCC on the same terms as any clinic or patient contributor — no preferential rate, no allocation bonus. Cleartext PHI workloads run only in attested TEEs.
Seven workload categories
| Workload | PHI scope | Eligible tiers |
|---|---|---|
| AI master model — public + synthetic Train on PubMed, MIMIC-IV, Synthea, public guidelines | None | Patient · Clinic · Corporate |
| AI master model — own-data federated Local training on contributor's own PHI; gradients leave under secure-aggregation + DP | Contributor own | Patient · Clinic · Corporate |
| AI master model — TEE cross-patient Other-patient cleartext PHI in attested enclave | Other-cleartext | Clinic-TEE · Corporate-TEE only |
| User-AI inference — knowledge class Knowledge-only queries (no PHI in query or response) | None | Any tier |
| User-AI inference — personalized class PHI-bearing queries; routes to user device or HIPAA-covered TEE | Contributor own / TEE | User device · Clinic-TEE · Corporate-TEE |
| Encrypted-data analytics Homomorphic / secure-MPC on encrypted PHI | Encrypted other-patients | Patient · Clinic · Corporate |
| Synthetic-data generation IRB-approved synthetic patient cohorts for research / model training | None | Any tier |
Anti-gaming controls
- TEE attestation — Intel SGX / AMD SEV / Apple Secure Enclave remote attestation; workload manifest required for any cleartext PHI workload
- Hardware binding — one wallet per attested hardware identifier; sybil-creating new wallets per machine doesn't multiply earnings
- Workload challenge-response — deterministic challenges verify the workload was actually executed
- Probabilistic audit replay — a fraction of mints are audit-replayed on alternate hardware; mismatches result in slashing
- Power-cost dashboard — published efficiency floors; contributor sees expected net of electricity before opting in
Getting started
Patient app: navigate to Settings → Compute Contribution to opt in. Hardware fingerprint is registered to your wallet; the protocol assigns workloads matched to your hardware capability + your opt-in scope + your energy preference. Earnings are visible in real-time on /portfolio.
Clinic / corporate: TEE-attestation onboarding is operated by the validator-registry service. Apply via /staking (validators page) and select "compute-attestation node" as the lane.
Disclaimer. Compute mining is a Phase-2 lane in the protocol roadmap; pre-launch the workload dispatch runs in test mode against synthetic data. HCC earned through compute contribution is reported on Form 1099-MISC per IRS Notice 2014-21. Patients should consult a tax advisor before opting in. See /mining for the full mining-lane taxonomy and /legal/cftc for the classification posture.