A small tagger that finds secrets and personal data in text before it leaves your machine, with the synthetic data we wrote for it.
AI & ML interests
on device models
Recent Activity
Organization Card
a one-person company and her agents (and help from her pet). we make small models that run on your own machine and do one job for an agent harness, and we publish what we trained them on.
web-extract · search and page-reading for agents
- egemma-td-v2: decides what's worth opening (272M, CPU)
- smallthink: an LFM2.5-24B that thinks for 66 tokens, not 2,000
egemma-pii-secrets · finds secrets and personal data in text before it leaves your box
every card says where the model fails, and every number says what it was measured on.
made by ana (@okami.mom) with bunny, bonsai and dawn · albedo.sh · wisp.place
models 4
nekomimi-llc/LFM2.5-24B-A2B-smallthink-lora
Text Generation • Updated • 48
nekomimi-llc/LFM2.5-24B-A2B-smallthink-GGUF
Text Generation • 24B • Updated • 309 • 1
nekomimi-llc/egemma-pii-secrets
Token Classification • Updated • 564 • 1
nekomimi-llc/egemma-td-v2
Feature Extraction • 0.3B • Updated • 79 • 2