Enable SPs to monetize idle machines with minimal setup by acting as API forwarders.
SPs deploy lightweight API servers on their idle machines to route client requests to inference models (hosted locally or on external compute nodes). Requires minimal compute resources, making it accessible for SPs with varying hardware capabilities.
Compute Resource Provision
SPs deploy inference models on their compute machines, supported by Filecoin’s compute orchestration tools (e.g., Bacalhau or Filecoin Virtual Machine). Models are executed locally, with SPs handling the full inference pipeline (request processing, model inference, response gener…
Storage and Distributed Training (Future)
Store inference models and request/response data on Filecoin, using Content Identifiers (CIDs) for verification. Enable SPs to contribute compute and storage resources for distributed training, allowing clients to train large models across SP machines. Smart contracts manage stor…
Filecoin ProPGF
This project creates a decentralized marketplace for AI model inference services, enabling Filecoin Storage Providers (SPs) to monetize their idle compute machines post-data encapsulation. By repurposing these underutilized resources, SPs can offer inference services compatible w…