EVALUATE. VERIFY.
mask.
mask enables buyers to evaluate dataset quality via confidential AI inference inside 0G Private Computer TEEs—without sellers exposing sensitive records.
Confidential Discovery via
Private AI Inference
Data discovery is broken by the privacy-confidence tradeoff. Sellers fear data exposure, while buyers fear purchasing blind. mask solves this paradox by combining confidential AI inference inside 0G Private Computer TEEs with cryptographic verification.
Buyers ask plain-language questions about completeness, accuracy, and formatting. Sellers upload a sample capped at 50 records. Inside the isolated TEE enclave, the AI model generates scored insights without ever storing or logging raw data.
How mask Works: The Evaluation Flow
A 5-step confidential analysis flow combining plain-language evaluation, 50-row bounded CSV sampling, 0G Private Computer TEE execution, and cryptographic verification.
Specify questions in plain language: Is this dataset complete? Are records up-to-date? No schemas or rubrics needed. mask creates a private buyer link for the seller.
Seller opens the private link and uploads a CSV sample (capped at 50 records). The sample is validated in memory and never stored on disk.
Sample and questions route through 0G Router to 0G Private Computer. Inference runs inside an isolated Trusted Execution Environment (TEE).
Response returns with full metadata: 0G Model ID, Provider Info, Request ID, Token usage, and TEE attestation proof. If inference fails, no score is published.
Armed with scored insights, confidence levels, metadata, and full auditability, the buyer makes an informed dataset purchasing decision.