MLOps Engineer (удаленная работа)

Вакансия: MLOps Engineer

Описание вакансии

We are looking for an experienced MLOps Engineer to work on the project - a decentralized AI protocol on Monad that leverages idle consumer hardware for swarm inference. It enables Small Language Models to achieve advanced multi-step reasoning at lower costs, surpassing the performance and scalability of leading models.

Responsibilities:

• Deploy scalable, production-ready ML services with optimized infrastructure and auto-scaling Kubernetes clusters, create Helm templates for rapid Kubernetes node deployment.

• Optimize GPU resources using MIG (Multi-Instance GPU) and NOS (Node Offloading System);

• Manage cloud storage (e.g., S3) to ensure high availability and performance.Deploy and manage large language models (LLM), small language models (SLM), and large multimodal models (LMM);

• Serve ML models using technologies like Triton Inference Server, optimize models with ONNX and TensorRT for efficient deployment;

• Set up monitoring and logging solutions using Grafana, Prometheus, Loki, Elasticsearch, and OpenSearch;

• Write and maintain CI/CD pipelines using GitHub Actions for seamless deployment processes.

Requirements:

• 5+ years of experience in MLOps or ML engineering roles;

• Proficiency in Kubernetes, Helm, and containerization technologies;

• Experience with GPU optimization (MIG, NOS) and cloud platforms (AWS, GCP, Azure);

• Strong knowledge of monitoring tools (Grafana, Prometheus) and scripting languages (Python, Bash);

• Hands-on experience with CI/CD tools and workflow management systems;

• Familiarity with Triton Inference Server, ONNX, and TensorRT for model serving and optimization.

As a plus:

• Bachelor s or Master s degree in Computer Science, Engineering, or a related field;

• Experience with advanced ML techniques, such as multi-sampling and dynamic temperatures;

• Knowledge of distributed training and large model fine-tuning;

• Proficiency in Go or Rust programming languages;

• Experience designing and implementing highly secure MLOps pipelines, including secure model deployment and data encryption.