Machine Learning Engineer

About the role

Neuralcraft builds AI for the rooms where it has to be defended- healthcare, finance, insurance, and the public sector. As a Machine Learning Engineer, you’ll take models from a notebook to a governed production system: trained on the right data, evaluated before every deploy, and explainable on every prediction.

This is a hands-on role for someone who enjoys the whole lifecycle- data, training, evaluation, serving, and the monitoring that keeps it honest once it’s live. You’ll work directly with our founders and our customers’ technical teams, often inside their private cloud.

What you’ll do

• Design, train, and fine-tune models- LLMs, classical ML, and retrieval pipelines- on customer-specific, often sensitive, distributions.
• Build the evaluation harness that gates every deploy: offline eval suites, online checks on live traffic, and drift/bias monitors.
• Ship per-prediction explainability so a model’s “why” is one click, not a three-week investigation.
• Stand up training and inference inside VPC / on-prem environments where data never leaves the perimeter.
• Generate and govern synthetic data where real data is restricted or unavailable.
• Partner with design and product to turn hard, regulated workflows into reliable systems.

What we’re looking for

• 3+ years building ML systems that reached production (not just experiments).
• Strong Python and the modern ML stack (PyTorch, Hugging Face, vector stores, an orchestration framework or two).
• Real experience with LLMs- fine-tuning, retrieval-augmented generation, and evaluation.
• A bias toward measurement: you reach for an eval before you reach for a bigger model.
• Comfort owning a problem end to end and communicating clearly with non-ML stakeholders.

Bonus points

• Worked in a regulated domain (HIPAA, SR 11-7, model governance) or alongside a compliance/audit team.
• Experience with private-cloud or on-prem deployment, MLOps, and observability tooling.
• Open-source contributions, papers, or side projects we can look at.

Why Neuralcraft

You’ll work on consequential problems with a small, senior team, ship to real users in weeks, and own the outcome- not just the model. We care about traceability, doing right by the people accountable for these systems, and moving at the speed of trust.