Machine Learning Scientist - AI Trainer

DataAnnotation; Madison (WI); Remote; USD 170,000 - 243,000; Full time

Benefits offered by this job

Remote work

Flexible schedule

Project-based work

Job summary

DataAnnotation is seeking a Machine Learning Engineer to stress-test how models reason about ML systems, including training dynamics, evaluation design, data pipelines, and deployment trade-offs. You will write prompts that probe model reasoning and review AI outputs for subtle errors, then craft the correct solution grounded in practitioner experience.

The role emphasizes hands-on work with Python and the modern ML stack (PyTorch/JAX).

Qualifications

• Hands-on experience training, evaluating, or deploying models professionally or in serious personal work.
• Comfort with the modern ML stack; tasks assume Python and PyTorch or JAX.
• Clear written English; explanations are the training signal.
• No degree required; focus on what you can do.

Responsibilities

• Write prompts that probe how models reason about training, evaluation, debugging, and productionizing ML systems.
• Review AI output for subtle errors: leaky evaluations, wrong loss formulations, and misdiagnosed training failures.
• Write the correct solution when the model falls short, grounded in real practitioner experience.

Skills

Python

PyTorch

JAX

Clear English writing

Tools

PyTorch

JAX

Git

Job description

Models are surprisingly bad at reasoning about themselves: training dynamics, evaluation design, data pipelines, and deployment trade-offs. Plausible-sounding ML advice is often subtly wrong.

As a Machine Learning Engineer you'll stress-test how models reason about ML systems and write the answers a strong practitioner would give, shaping how the next generation handles your field.

What you’ll actually do

• Write prompts that probe how models reason about training, evaluation, debugging, and productionizing ML systems.
• Review AI output for subtle errors: leaky evaluations, wrong loss formulations, and misdiagnosed training failures.
• Write the correct solution when the model falls short, grounded in real practitioner experience.

What we look for

• Hands-on experience training, evaluating, or deploying models professionally or in serious personal work.
• Comfort with the modern ML stack; most tasks assume Python and PyTorch or JAX.
• Clear written English: your explanations are the training signal.
• No degree required. We care about what you can do, not where you learned it.

Compensation

Up to $150+/hr depending on task difficulty and specialization. Many contributors add $100k–$200k a year; some make it their full-time income.

About DataAnnotation

DataAnnotation is where 1M+ experts train the world’s leading AI models. Hundreds of millions paid to contributors to date, and the average contributor stays 5+ years. Flexible, remote, and always project-available.