Pluralis Research

Research Scientist

• Build the RL post-training stack end-to-end, including rollout ingestion, reward computation, policy updates, and returning updated weights to the network
• Set the direction for the post-training stack and execute its development
• Adapt RL algorithms to asynchronous, high-latency, partially trusted generation
• Address staleness tolerance, off-policy corrections, and communication-efficient policy updates
• Build evaluations demonstrating model improvement
• Deliver the first decentralized post-trained model from run to public artifact

Explore Research Scientist – Post-Training at Pluralis Research. Review requirements and application details.

Research Scientist – Post-Training

• Build and own the inference stack end-to-end
• Set direction for pipeline-parallel execution, placement and routing, transport, serving engine, and failure handling
• Invent, design, validate, and productionize algorithms for fast inference on consumer hardware over the public internet
• Keep the rollout pipeline fast and reliable for reinforcement learning training
• Develop the serving layer for models once trained
• Support inference across Macs and consumer GPUs owned by distributed participants
• Handle nodes joining and leaving mid-run and model weights changing during training

Explore Research Scientist – Geo-Distributed Inference at Pluralis Research. Review requirements and application details.

Research Scientist – Geo-Distributed Inference

• Build the training system that takes Protocol Learning from an 8B run to frontier scale
• Implement and optimize model-parallel distributed pretraining across heterogeneous GPUs
• Implement data, pipeline, and tensor parallelism over low-bandwidth, high-latency links
• Reduce communication overhead while maintaining model convergence
• Build elasticity and fault tolerance so runs survive node churn
• Implement robust checkpointing, state synchronization, and recovery as participants join and leave
• Build monitoring to show throughput, bottlenecks, and model quality across hundreds of devices
• Train and serve large models on heterogeneous hardware in physically different regions over ordinary internet connections

Explore Research Scientist – Pre-training at Pluralis Research. Review requirements and application details.

Research Scientist – Pre-training

• Enumerate and maintain the threat model for malicious or careless workers across pre-training, post-training, and inference
• Assess training disruption, denial-of-service, free-riding, model poisoning, backdoors, data extraction, reputation manipulation, and reward manipulation
• Build statistical verification methods with stated error rates
• Tune verification tests using rigorous benchmarks
• Control false positives and false negatives across heterogeneous hardware, including different GPUs and Macs
• Build and operate the verification service in the inference path
• Verify that tokens came from the claimed model and sampling parameters
• Verify that training contributions are what they claim to be

Explore Research Scientist – Decentralized Training and Inference… at Pluralis Research. Review requirements and application details.

Research Scientist – Post-Training

Research Scientist – Geo-Distributed Inference

Research Scientist – Pre-training

Research Scientist – Decentralized Training and Inference Verification