Data science and machine learning AI training jobs: tasks, pay and who qualifies

What data science and ML AI training work involves

AI models are increasingly asked to analyse data, build models and write up results, and they make the mistakes a junior analyst makes. Data science and machine learning AI training jobs pay practitioners to catch those mistakes and to build the tasks that expose them. On 29 September 2026 this site listed 36 roles under Data, AI & ML, out of 699 listings in total.

The listings describe four kinds of work.

Evaluating analysis. micro1's Data Scientist listing asks you to judge model outputs on statistics, ML and quantitative problems, create reference solutions and catch flawed methodology: a leaky validation split, a p-value read backwards, a confounded A/B test. Mercor's Data Science Expert listing does the same for analysis decks and experiment write-ups, at $100–150.

Building ML engineering tasks. micro1's ML Engineer creates and solves hard tasks such as implementing a model component, building a reproducible training workflow or debugging a numerical failure, at $100–150.

Auditing tasks. Mercor's ML Challenge Task Auditor checks applied ML tasks for sound experiment design, honest splits, data leakage and metric gaming, at $70–90.

Chart reasoning benchmarks. micro1's statistical chart analysis listing writes questions that force a model to reason from Sankey diagrams, calibration curves and residual plots rather than read a value off an axis, at $25–50.

What it pays

30 of the 36 quote an hourly rate. Two of those are facial data collection listings, which pay a one-off sum for photos rather than analytical work. The other 28 run from $25 to $350, with a median midpoint of $85 an hour. The top belongs to micro1's data scientist role at $245–280 and its executive presentation review listing at $150–350. Chart analysis and data analyst roles sit at $25–60.

Four micro1 listings are salaried Member of Technical Staff posts at $200,000–320,000 a year, and two Mercor agent engineering listings pay per task.

Who qualifies

Most listings ask for experience at a named level of employer. The micro1 data scientist listing asks for at least a year at a top company in tech, finance or research; Mercor's data science expert for five years at a top firm in the US, UK, Canada, Australia or New Zealand. The ML engineering roles are stricter on education: micro1's ML engineer asks for a master's or PhD and working experience with at least two frameworks, naming PyTorch, JAX and Hugging Face Transformers among them. Python is the most common skill across the 36 listings.

Entry is easier on the chart analysis roles, which accept a bachelor's degree or equivalent work plus two years with statistical models or visualisations. micro1 posts 21 of these roles and Mercor 15; 28 are open worldwide.

Heavier software work, such as GPU kernels, RL environments and coding benchmarks, is covered in our guide for software engineers.

Where to start

Browse the data, AI and ML filter. Statisticians and quantitative scientists should also check our STEM guide.