Apply on Employer Site

Opentensor Foundation · 3 days ago

Research Engineer – Distributed Training

United States

Full-time

Remote

Mid, Senior Level

Opentensor Foundation is building the infrastructure for decentralized AI at internet scale, supporting the core development of the Opentensor Foundation and the broader Bittensor ecosystem. The Research Engineer will help design scalable training solutions for decentralized AI model training, focusing on innovative research efforts and enhancing AI workload efficiency.

Artificial Intelligence (AI)BlockchainInformation TechnologySoftware

Hiring Manager

Ryan Staab

Responsibilities

Drive innovative research efforts focused on building a large-scale, secure, and dependable system for orchestrating decentralized AI model training

Continuously refine and enhance AI workload efficiency by applying cutting-edge techniques in compute and memory optimization

Actively contribute to shaping our open-source tools and libraries that support scalable, distributed training of machine learning models

Share breakthroughs and research findings with the broader community through publications in premier conferences like NeurIPS

Keep a close eye on emerging trends in ML infrastructure, decentralized compute, and tooling—proactively identifying ways to evolve and improve the platform’s performance and developer experience

Qualification

Distributed trainingAIMachine learningMLOps workflowsPyTorch DistributedDeepSpeedRayMosaicML's LLM FoundryExperiment loggingCI/CD processesParallelism strategies

Required

Proven expertise in AI and machine learning engineering, with a track record of building and scaling complete pipelines for training and deploying large-scale AI models

In-depth knowledge of distributed training methodologies and frameworks such as PyTorch Distributed, DeepSpeed, Ray, and MosaicML's LLM Foundry, with a focus on enhancing training efficiency and system scalability

Hands-on experience training large models at scale, leveraging advanced parallelism strategies including data, tensor, and pipeline parallel techniques

Strong grasp of modern MLOps workflows, including model lifecycle management, experiment logging, and automation through CI/CD processes

Deeply motivated by the mission to push the boundaries of decentralized AI training and make cutting-edge AI technology more accessible to a global community of developers and researchers