AI-05 · AI infrastructure · Open
Senior AI Infrastructure Engineer
Build ErisAI as a credible on-chain AI runtime first: training, evaluation and inference jobs that produce verifiable receipts.
- Location
- Remote
- Engagement
- Contract-to-hire
- Path
- Specialist pathway
Mandate
Own the model runtime, GPU-subnet scheduling, determinism boundaries and telemetry for the first ErisAI compute slice.
ErisAI runs training and inference as on-chain contracts. Heavy compute stays on GPU subnets, and every job must produce a receipt the chain can verify before the runtime earns a wider scope.
First 90 days
Four outcomes that turn the thesis into a running system.
- 01
Deliver a documented on-chain inference slice with a verifiable receipt.
- 02
Implement deterministic job execution and model-loading fixtures.
- 03
Define telemetry for latency, cost, accuracy and abuse signals.
- 04
Specify the narrow conditions where GPU-subnet training improves the product.
You will own
- Design the model runtime and job-execution architecture.
- Model training, evaluation, inference and receipt flows.
- Integrate GPU-subnet scheduling where compute demands it.
- Create evaluation harnesses, telemetry and anti-abuse controls.
- Collaborate on provenance and verifiable-compute experiments.
Strong fit
- Shipped ML systems or inference-runtime experience.
- Strong Python, TypeScript and WebAssembly skills.
- Model-serving, ONNX and evaluation literacy.
- Skepticism about unnecessary blockchain mechanics.
- Ability to prototype without leaving fragile architecture.
Stack
Application
Show how telemetry changed a runtime or evaluation decision.
Share public or sanitized material only. Do not send confidential source code, private keys or client-restricted documents.
- An ML or inference system you owned
- A determinism or evaluation decision driven by telemetry
- A compute path you deliberately kept off-chain
