1 of 7

Lit Data Market

AI Inference over Private Data

Nelson

2 of 7

The Problem

  • Hard to license data without giving access to the data
    • Can provide results but requires locking inference to the provider
  • Privacy for data sources
  • Transparency on data access and requests

3 of 7

Data Market - Unlocking Inference with Data

Users can run inference on data without seeing the full data

Privacy

Optional signed paper trail of who made what request

Prevent abuse of data (eg: spying on their spouse)

Transparent

Data vendors can charge for access to data without giving up the full data

Monetize

4 of 7

How It Works

  1. Users connect with their Eth wallet to Lit nodes
  2. Users make LLM queries
  3. LIT decrypts API keys based on user wallet
    1. API key and credit tracking for data vendor
    2. API key and credit tracking for inference vendor
  4. LIT action takes data from data vendor and uses it with inference vendor. User doesn’t see the input data
  5. LIT action returns results to user and signs that user made that request
    • Query is tracked, but not the data or the result, maintaining privacy with accountability

5 of 7

Demo

6 of 7

Improvements for Future

  • Improved monetization
    • Marketplace for data vendors and inference vendors
  • Improved UX
  • Improved access controls
    • Restricting to specific wallets, tying credit balances to wallets, etc
  • Improved interoperability
    • ​

7 of 7

Team

Nelson Lai

Software Engineer

Likes Chinese food�Learning about AI and crypto.

​

​