Skip to main content
Every time you send a request to Pinaivu, two distinct actors work behind the scenes to deliver your response: the Coordinator, which orchestrates the process with hardware-backed trust, and a network of independent GPU nodes that compete to serve your inference. You never interact with either directly — but understanding what they do explains why Pinaivu is more reliable and verifiable than a traditional centralized API.

​
The Coordinator

The Coordinator is the trust anchor of the Pinaivu network. It runs inside an AWS Nitro Enclave — a hardware-isolated, cryptographically attested environment. This means the code running inside the Coordinator can be verified by anyone, and no one — not even Pinaivu — can read its memory, alter its logic, or tamper with its decisions once it’s deployed. The Coordinator is responsible for the full lifecycle of every inference request:
  • Receiving requests forwarded from the gateway
  • Running auctions by broadcasting each job to available GPU nodes across the decentralized network
  • Dispatching the winning node based on bid criteria
  • Settling payments to the node that served the inference
  • Signing routing receipts that permanently record the outcome
Because the Coordinator is attested, its routing decisions and signed receipts carry cryptographic weight. You don’t have to trust Pinaivu’s claims about how a request was handled — the signed receipt proves it.

​
GPU Nodes

GPU nodes are independent operators who have connected their hardware to the Pinaivu network. They range from individual hobbyists to professional compute providers, all participating in the same open marketplace. When a request arrives, every available node can submit a bid. The Coordinator evaluates those bids and selects the winner — typically the node offering the best combination of latency and cost. This competitive dynamic has real consequences for you as a user:
  • Low latency: nodes are incentivized to respond quickly, since slow bids lose to faster ones.
  • Fair pricing: no single operator can charge arbitrarily — they’re always competing against peers.
  • Resilience: if any node goes offline or behaves poorly, others immediately fill the gap in the next auction.
Nodes communicate with the Coordinator over a decentralized peer-to-peer protocol — there is no central message broker in the path.

​
What this means for you

  • Verifiable routing: every request produces a signed receipt you can inspect on the Pinaivu Explorer. You can confirm which node served your inference without taking anyone’s word for it.
  • Competitive performance: market-driven bidding means you benefit from nodes actively optimizing for speed and availability.
  • No vendor lock-in on compute: Pinaivu doesn’t own or operate the GPU nodes. The network is open to any operator, which means capacity and competition grow independently of Pinaivu itself.
You don’t need to interact with nodes or the Coordinator directly. The gateway at https://api.pinaivu.com/v1 handles all of this automatically on every request.