Research Scientist - Post-Training
AI
Pluralis Research is hiring a Research Scientist for post-training to build an RL post-training stack for decentralized training of large models on consumer GPUs and Macs over the public internet. The role covers end-to-end RL training loops, novel algorithms for asynchronous and partially trusted generation, and shipping the first decentralized post-trained models. Requires hands-on RL post-training experience on LLMs and strong production-quality Python and PyTorch engineering.
What you'll do
- Build the RL training loop end-to-end: rollout ingestion from the geo-distributed inference pipeline, reward computation, policy updates, and getting updated weights back out to the network
- Adapt standard RL recipes to asynchronous, high-latency, partially trusted generation: staleness tolerance, off-policy corrections, and communication-efficient policy updates
- Build the evals that show the models are improving
- Take the first decentralized post-trained release from run to public artifact
What you need
Must-have
Reinforcement LearningRLHFRLVRReasoning-focused RLRollout generationAsync training loopsWeight synchronizationPython
Nice-to-have
Slow network trainingDecentralized setupsFederated setupsvLLMSGLangReward modelingVerifiable-reward datasetsP2P networkingNAT traversal
Soft skills
Mission alignmentResearch abilityOwnershipComfort working across timezonesProfessional-level English proficiency
Research ScientistPost-TrainingReinforcement LearningDecentralized AIProtocol LearningPyTorchLLMMachine LearningHybridAustralia