Machine Learning Engineer, MLOps

Looking for a Machine Learning Engineer, MLOps who wants to stay somewhere for a while. We're looking for an engineer to build retrieval-augmented generation features on hosted and open-weight language models, with proper evaluation. Remote position. Most of the team is in Berlin, but nobody is required to come in.

What you'll be doing:

• Fine-tune open-weight models with LoRA when it beats prompting.
• Cut inference cost with quantization, batching and caching.
• Track experiments, artifacts and versions in MLflow or Weights & Biases.
• Orchestrate pipelines with Kubeflow, SageMaker Pipelines, Vertex AI or Airflow.

Requirements:

• Hands-on experience with PyTorch, TensorFlow or JAX.
• Knowledge of LLM APIs, prompt design and evaluation methods.
• BS or MS in computer science, engineering or a related field.
• Strong Python and solid engineering habits (testing, packaging, code review)

Nice to have:

• Published papers or open source ML contributions.
• Distributed training with DeepSpeed or FSDP.

Pay and benefits:

• Company pension scheme.
• 30 days of annual leave.
• Childcare allowance.
• Home office equipment.

About Marblebound Labs:

Annual releases gave way to two-week sprints years ago, and the platform now handles over 200 million API calls a day. Marblebound Labs employs 700 people across engineering hubs in Berlin and three other locations, building ERP software for mid-market manufacturers. Most engineers work remotely and come into the office one week a quarter for planning.

Shortlisted candidates will hear from us within two weeks.

Please mention your notice period and expected salary.

Start: Immediately.

Demo job created on 11jobs for testing and demonstrations. It is not an actual opening at Marblebound Labs.