Machine Learning Engineer

Role overview Own the machine-learning work behind custom systems for business operations, including document extraction, retrieval, classification, and assistants that use tools. You’ll shape the problem with clients and engineers, choose or adapt suitable models, measure performance, and keep improving quality after launch. Delivery typically runs in approximately six-week cycles; the work focuses on applied systems rather than training foundation models.

Responsibilities - Define the task, inputs, outputs, and constraints such as accuracy, latency, and cost with clients and the engineering team. - Design and ship retrieval, extraction, classification, ranking, and tool-calling systems, selecting classical ML when it is the better fit. - Build offline evaluation datasets and harnesses to measure quality, speed, and cost before release. - Integrate existing models or train, fine-tune, or distill models when hosted APIs do not meet the need. - Integrate models into production services in Python and contribute across services written in JavaScript or TypeScript. - Monitor live systems, investigate failures, add regression coverage, and guide subsequent improvements.

Requirements - At least three years of experience shipping machine-learning systems in production. - Strong Python skills and the ability to read and write JavaScript or TypeScript. - Experience with PyTorch, scikit-learn, or a comparable training and evaluation toolkit. - Ability to evaluate model quality, including accuracy, precision and recall, hallucinations, latency, and cost, and use findings to improve the system. - Availability to overlap some working hours with teams in Sydney, Austin, or Los Angeles.

Nice to have - Production experience with retrieval-augmented generation, tool calling, structured extraction, or evaluation harnesses. - Model fine-tuning or distillation experience. - Monitoring and incident response for products that use models. - Experience turning operational needs into clear specifications and datasets with non-engineering colleagues.

Benefits and work setup - Full-time, remote work across Australia and the United States, with some timezone overlap required.