Description

• Engineer who will build and operate our LLM fine-tuning and model-serving platform while also acting as an internal AI consultant, helping other teams take their use cases from idea to a working fine-tuned or deployed model. The ideal candidate combines deep hands-on fine-tuning expertise with strong distributed-systems engineering, and is equally comfortable debugging a training run and advising a non-ML team on how to frame their problem.

Responsibilities

• Design, build, and maintain fine-tuning pipelines (LoRA and full fine-tuning) for both generative and classification tasks
• Own model training reliability: precision/memory issues, convergence problems, evaluation metrics, early stopping, and reproducibility across environments
• Manage the handoff from training to serving, including quantization and inference-time configuration
• Build and maintain the underlying platform infrastructure (containerized training/serving workloads on Kubernetes, GPU scheduling, observability)
• Act as embedded consultant for other teams: review their use case, recommend model/approach, and troubleshoot underperforming or broken fine-tunes
• Diagnose issues that span both data/ML quality and system/infra layers, distinguishing which is the actual root cause
• Contribute to technical design docs, architecture decisions, and code reviews with an ML-aware lens
• Produce lightweight documentation and enablement material so consulting scales beyond one-on-one support

Qualifications

• Hands-on experience fine-tuning LLMs using PyTorch and common libraries (e.g., HuggingFace transformers, PEFT, datasets)
• Practical experience with model serving and quantization, and how training-time vs. inference-time behavior differs
• Strong Kubernetes fundamentals (containers, scheduling, resource management on GPU infrastructure)
• Senior-level Python skills and ability to read across a full-stack codebase to trace an issue
• Strong communication and consulting skills — able to translate another team's use case into a technical plan
• Preferred: experience with both generative and classification fine-tuning tasks, GPU observability/cost tuning, and multi-environment ML platforms

Key Skills

Education

Any Graduate