Senfin-A

Our project provides a simple process for generating fake data from real information while keeping personal details hidden. A user first supplies real data, which is processed in a secure AI environment using Nillion SecretLLM. This environment reads and learns from the data without revealing any sensitive information. The AI then creates a synthetic version of the data that preserves important patterns and statistics but removes any personal or identifying details.

Both the synthetic output and the AI's chain of thought logs are saved in Recall, a blockchain based storage system where everything is recorded in a single agent index bucket for easy tracking and retrieval. Anyone with permission can later look up or download the synthetic dataset and, if desired, the AI's thought logs. This approach provides a safe and privacy respecting way to explore or analyze data without exposing real personal information.

Our system is powered by a combination of Nillion’s SecretLLM technology, which handles private AI generation inside a secure trusted execution environment, and Recall’s blockchain based data storage that ensures tamper proof recording of both synthetic outputs and AI reasoning logs. On the backend, we have a Node.js agent that receives real data or references from the user, sends it through SecretLLM to produce a privacy preserving synthetic dataset, and then saves the result in a single agent index bucket on Recall. The chain of thought logs, which capture the AI's behind the scenes reasoning, are also stored in the same Recall bucket, so everything about a given generation can be traced back to a unique timestamp or ID. We manage the agent's private key and token credits in .env files, using Recall's SDK and credit purchase logic to ensure the agent can store data without leaking any keys or personal info. In addition, we run a simple gating mechanism that can be extended for user authentication, payment checks, or token based access, ensuring only authorized individuals can fetch or download the records. Everything else runs on a Node Express server that exposes routes for generating synthetic data, listing stored objects, and retrieving them by key.