MLOps Engineer

About the role

The MLOps Engineer will define and own the machine‑learning infrastructure strategy for AI71’s platform, covering both SaaS and on‑premise deployments. You will shape how large language models and other deep‑learning models are deployed, fine‑tuned and served at scale, while mentoring engineers and driving multi‑quarter infrastructure road‑maps.

Key responsibilities

• Design the end‑to‑end ML infrastructure architecture, including model deployment (vLLM, Triton, TGI) and pipeline tooling (MLflow, Kubeflow) across AWS, Azure and GCP.
• Set reliability targets, implement monitoring, and lead incident response for research and production environments.
• Guide cross‑team initiatives to improve inference performance and cost efficiency using distributed training frameworks such as DeepSpeed, FSDP and Accelerate.
• Partner with ML researchers, product leaders and engineering management to define multi‑quarter infrastructure strategy and ensure scalability for both managed SaaS and air‑gapped on‑prem deployments.

Required profile

• 10+ years of experience in MLOps, ML infrastructure or machine‑learning engineering with a proven record of architectural ownership.
• Demonstrated ability to mentor senior engineers and raise operational standards across multiple teams.
• Strong communication and stakeholder‑management skills, with a commitment to building diverse and inclusive engineering teams.

Required skills

• Python programming
• Cloud platforms: AWS, Azure, GCP
• Kubernetes (GPU scheduling, multi‑tenancy, operators)
• Model serving frameworks: vLLM, Triton, TGI
• Pipeline orchestration: MLflow, Kubeflow
• Distributed training: DeepSpeed, FSDP, Accelerate
• GPU systems: CUDA, NCCL, NVLink, InfiniBand/RoCE

What we offer

• Mission‑driven work on cutting‑edge AI applications that address real‑world challenges.
• Competitive compensation, benefits and clear career‑growth pathways as a foundational team member.
• Flexible working environment with access to the latest tools and technologies.