AI & ML interests

Retrieval、Embeddings、Reranking、Agent memory、ONNX、Quantization

Recent Activity

Organization Card

Påmin Labs

We build Påmin Memory: universal memory for AI agents. It keeps durable evidence, versioned knowledge, hybrid retrieval, temporal context and explainable recall in one local-first engine.

The models we publish here are the ones Påmin Memory runs. Each model card states:

  • what was changed from the original model;
  • how faithful the result is, measured rather than assumed;
  • how to rebuild it from the upstream release.

Models

Our optimized int8 ONNX build of BGE-M3 for CPU is in preparation and will be published here with its build pipeline and measured fidelity.

Links

datasets 0

None public yet