Pluralis Research
Machine Learning Engineer - ML Training Platform
San Francisco
Sponsorship not specifiedDetected 112 days ago
PythonNode.jsDistributed SystemsAWSGCPAzureCloud PlatformsDockerKubernetesTerraformPrometheusGrafanaSite Reliability EngineeringPlatform EngineeringMachine LearningA/B TestingResearch
About the role
- Pluralis Research https://pluralis.ai/ is pioneering Protocol Learning https://arxiv.org/abs/2412.07890-a fully decentralised way to train and deploy AI models that opens this layer to individuals rather than well resourced corporates.
- By pooling compute from many participants, incentivising their efforts, and preventing any single party from controlling a model's full weights, we're creating a genuinely open, collaborative path to frontier-scale AI.
Responsibilities
- Multi-Cloud Infrastructure: Design resource management systems provisioning and orchestrating compute across AWS, GCP, and Azure using infrastructure-as-code (Pulumi/Terraform).
- Real-World Networking: Build systems that simulate and handle real-world network conditions - bandwidth shaping, latency injection, packet loss - while managing dynamic node churn and ensuring efficient data flow across workers with heterogeneous connectivity, because our training happens on consumer nodes and non co-located infrastructure, not in a datacenter.
Requirements
- Experience in a startup environment with an emphasis on micro-services orchestration or big tech background
- Ideally, you'll have 5+ years of work experience with deep experience in:
Skills
- Handle dynamic scaling, state synchronization, and concurrent operations across hundreds of heterogeneous nodes.
Benefits
- GPU clusters, NVIDIA runtime, S3 checkpointing, Large dataset management and streaming, health monitoring, and resilient retry strategies.
Apply directly at Pluralis Research →Create a free account for alerts like thisView Pluralis Research immigration profile
This listing is sourced directly from Pluralis Research's careers page and normalized into a canonical job model.