Machine Learning Expert (Remote, Part-Time)

What You'll Do6

• Write a modeling problem from a system you shipped, covering the dataset, objective, constraints, and the reference answer that anchors grading of a model's attempt.
• Review training and evaluation code that models produce, checking whether the loss matches the objective, whether the eval set leaks, and whether an ablation supports the stated conclusion.
• Flag model proposals that would train without error but fail in production when distribution shift, label noise, or a misleading metric enters the picture.
• Explain your reasoning in writing for every judgment, because project teams rely on that written record.
• Raise unclear instructions or ambiguous task specs instead of guessing what a client wants.
• Grade model attempts against the rubric you helped define.

Requirements5

• Professional experience training, evaluating, or deploying machine learning systems.
• Strong written communication, since most work requires explaining technical reasoning to others.
• Comfort with ambiguous tasks and a habit of flagging unclear instructions.
• A track record of putting models in front of real users and dealing with the results.
• Where you trained carries less weight than evidence that you have handled real production feedback.

Who Should Apply

People who fit this pool have trained, evaluated, or deployed machine learning systems in a professional setting and can write clear explanations of their technical choices. They handle incomplete task specs with patience and ask questions when instructions leave room for doubt. The role suits practitioners who want part-time, project-based AI training and evaluation work rather than a fixed full-time job. Candidates who need a defined start date or a single client assignment may find the standing listing frustrating, since no immediate opening is guaranteed. Weak written reasoning and a lack of production experience are the most common reasons applicants score low for this kind of work, and vague claims about model work without shipped examples also hurt.

Salary Insight