Machine Learning Engineer

Most of the AI work that reaches us isn’t “build a model from scratch” — it’s “make an existing workflow faster or cheaper using the model that actually fits,” which is a different and, we’d argue, more useful skill. You’ll scope those calls and build the systems around them.

What you’ll do

• Evaluate whether a client problem needs a custom model, a fine-tuned open-weight model, or a well-integrated call to an existing API — and make the case for whichever is actually right, including “none of the above”
• Build and productionize data pipelines that feed ML systems — the unglamorous 80% of the work that determines whether the model performs in production the way it did in the notebook
• Own model evaluation: define the metrics that map to the client’s actual business outcome, not just accuracy on a holdout set
• Integrate models into existing application backends — batching, caching, fallback behavior when a model call fails or times out
• Monitor deployed models for drift and degradation, and know when a model needs retraining versus when the upstream data pipeline is the actual problem

What we’re looking for

• 3+ years shipping ML systems to production, not just research or Kaggle work
• Strong Python fundamentals and comfort with at least one ML framework (PyTorch, TensorFlow) plus the surrounding tooling (pandas, a vector store, an orchestration tool)
• Experience with LLM-based systems — prompting, fine-tuning, or RAG architectures — since a meaningful share of current client work runs through them
• Ability to explain, in plain language, why a model is or isn’t the right tool for a given problem
• Bonus: experience with MLOps tooling (MLflow, Weights & Biases, or an equivalent) and cloud ML infrastructure

What you get

Varied problems instead of one product roadmap, the latitude to recommend a simpler non-ML solution when that’s the honest answer, and a team that won’t ask you to force-fit a model where a rules engine would do.