Discover everything.
Expose nothing.
Exemem is a privacy-preserving network over vectorized data. Search across everyone's data — without learning who owns it, linking entries to each other, or exposing raw content.
Fingerprints, not files
Every file becomes a vector embedding — a numerical fingerprint that captures meaning without revealing content. Fingerprints are public. Raw data stays private.
One disguise per entry
Each entry gets a unique derived pseudonym. No two entries can be linked to the same owner. Even 1,000 entries from one person look like 1,000 unrelated identities.
More users = stronger privacy
Network growth increases the anonymity set. Bigger crowd, harder to deanonymize. The opposite of traditional social networks.
Two layers
Public discovery. Private content. Separated by design.
Store your data
Upload files. AI generates vector embeddings — numerical fingerprints published to the public layer. Raw data stays behind Fold DB's access controls.
Discover across the network
Search all public embeddings with a similarity query. Results show per-entry pseudonyms and similarity scores — never real identities or linked entries.
Request access
Found a match? Request raw data through the entry's pseudonym. Fold DB enforces the owner's policies: trust distance, cryptographic keys, payment gates.
Applications
Privacy-preserving similarity discovery across any data type.
Photo Discovery
Search face embeddings across the network. Find 14 matching entries, each under a unique pseudonym. You can't tell if they belong to 14 people or 1. Request access or removal through the pseudonym.
IP Discovery
Embed your creative work — audio, text, code. Find entries with 96% similarity uploaded before your release. File a timestamped claim in the append-only store without knowing who owns the match.
Medical Cohorts
Patients publish clinical profile embeddings. Researchers discover 47 matching entries for a rare disease signature. Contact pseudonyms to propose trials — nobody's identity revealed unless they choose.
Threat Intelligence
Security teams publish embeddings of indicators of compromise. Search finds 7 matching entries. Share defenses through pseudonyms without revealing which organizations were hit.
Public discovery. Private content.
Two layers with different trust models. Discovery is open. Access is controlled.
| Public Layer | Private Layer | |
|---|---|---|
| What | Embeddings + Pseudonyms | Raw Data + Fold DB |
| Visibility | Open to everyone, no access gates | Policy-gated per entry |
| Content | Vector fingerprint (irreversible) | Original file (photos, docs, audio) |
| Identity | Unlinkable per-entry pseudonym | Owner verified via master key |
| Access control | None — embeddings are public | Trust distance, cryptographic keys, payment |
| Encryption | N/A (public data) | AES-256-GCM at rest |
| Index | Global ANN (HNSW/IVF), sublinear | Per-owner Fold DB instance |
| Store | Append-only, immutable record | Owner-controlled, deletable |
| Purpose | Discovery — find what exists | Access — control who sees it |
File
→
T_embed
→
Vector
→
Public Index
Access Request
→
Pseudonym Route
→
Fold DB Policy
→
Raw Data
Coming Soon
Exemem is in active development. Every new participant will strengthen privacy and improve discovery for everyone.