Principal AI-ML Infrastructure & Training Engineer

Audiience™ is transforming how content is created and trusted in publishing through innovative technology. The role involves architecting and owning the end-to-end ML training infrastructure, building scalable training pipelines, and collaborating with research teams to enhance production-ready pipelines.

Responsibilities:

• Architect and own the end-to-end ML training infrastructure - from data ingestion through experiment tracking to model checkpointing
• Build scalable, reproducible training pipelines that empower the research team to iterate fast without chaos
• Own compute orchestration, distributed training setups, and GPU cluster management
• Implement and manage experiment tracking (W&B, MLflow) and version-controlled data pipelines
• Collaborate directly with research to harden experimental approaches into production-ready pipelines
• Identify and systematically eliminate bottlenecks in training speed, cost, and reliability

Requirements:

• Deep experience with distributed training frameworks (FSDP, DeepSpeed, Megatron, or equivalent)
• Strong proficiency in PyTorch and modern ML tooling
• Experience with cloud compute orchestration (AWS, GCP, or Azure) at training scale
• Familiarity with containerization (Docker, Kubernetes) for ML workloads
• Solid understanding of ML experiment tracking, data versioning, and reproducibility
• Ability to profile and optimize training throughput and resource utilization
• Communication excellence – Can clearly articulate infrastructure decisions, tradeoffs, and failure post-mortems in writing
• Demonstrated ability to document systems architecture and explain technical decisions to a cross-functional team
• Prior experience as an ML infrastructure engineer, MLOps engineer, or applied researcher with strong infra instincts
• Startup or fast-moving environment experience is a plus
• Problem-solving prowess – You see problems others don't and solve them in ways others can't
• Tenacious learner – Self-taught capabilities and continuous improvement are in your DNA
• Systems thinker – You understand how complex systems interact and create elegant solutions
• Results-oriented – Bias toward flexibility, impact, and getting it done
• Collaborative by nature – You believe we can only win if we do it together
• Experience with CUDA, fused kernels, or low-level performance optimization
• Prior work with LLMs or large-scale foundation model training
• Experience building internal ML platforms or developer tooling for research teams
• Open-source contributions or published engineering writeups
• Previous startup or early-stage engineering experience
• Volunteer work