Research Scientist - Decentralized Training and Inference Verification

What you'll do

• Maintain threat models covering attacks on pre-training, post-training, and inference.
• Design and benchmark statistical verification tests, controlling false positives and false negatives across heterogeneous hardware.
• Build and operate the verification service in the inference path.

What they ask for

• Experience building a calibrated statistical decision system with stated error rates, or relevant publications in inference or training verification.
• Deep expertise in statistics and probability, including experiment design and decision-threshold calibration.
• Knowledge of verification approaches for untrusted compute, including statistical testing, re-execution, cryptographic proofs, and trusted hardware.
• Professional-level written and spoken English; comfortable working across time zones.

Nice to have

• Familiarity with large-scale pre-training and RL post-training.
• Familiarity with decentralized ML security and adversarial threat models, such as poisoning, Sybil, collusion, and replay.

Tools

GPUs, Macs

AI focus

Verification of decentralized AI training contributions and inference outputs, including detecting attacks and validating claimed models and sampling parameters.