AI training jobs for coders

372 live listings in Software Engineering, stating up to $300/hr.

Median top rate across the 255 hourly coders listings: $90/hr. Highest: $300/hr.

Why coders?

AI labs need humans who can write production-quality code, review AI-generated code for bugs, explain refactor decisions, and grade automated agent outputs for correctness.

What the work actually looks like

  • Reviewing AI-generated code for bugs, style, and correctness
  • Writing reference solutions for prompts the model needs to learn from
  • Grading multi-step agent runs that read/edit/run code
  • Refactoring legacy code with the model as a pair, then scoring its suggestions
  • Authoring difficult coding problems for benchmarking and RL training

Live listings for coders right now

Top 12 of 372 active listings, sorted by referral value + pay rate.

Turn open source maintenance judgment into stronger AI code

Build difficult AI tests for kernels, profiling and LLM serving

$75 - $110 per hour

Test AI reasoning about Kubernetes, AWS and infrastructure code

$70 - $110 per hour

Evaluate AI against real materials research and engineering

$70 - $110 per hour

Review and build Software Engineer tasks with clear feedback

$65 - $105 per hour

Challenge AI with production engineering judgment

Software Engineering

Mechanical Engineer

$60 - $90 per hour

Review mechanical designs to sharpen AI engineering judgment

$70 - $90 per hour

Audit complete AI assisted coding sessions for correctness

Software Engineering

SWE-Bench Task Auditor

$70 - $90 per hour

Audit repository benchmarks, reference patches and grading

Software Engineering

ML Challenge Task Auditor

$70 - $90 per hour

Audit applied ML experiments for rigor and reproducibility

$70 - $90 per hour

Audit AWS serverless designs and infrastructure code

Software Engineering

Kubernetes Task Auditor

$70 - $90 per hour

Audit Kubernetes tasks against production cluster behaviour

See all 372 listings →

What to expect (the honest version)

  1. The high-pay queue is gated behind a real coding test — usually a 60-90 minute screen with edge cases.
  2. Pay scales with language depth. Python + JS pay the entry-tier rates; deep C++, Rust, embedded, or low-level systems work pays more.
  3. Much of the coding evaluation work is queue-based and async rather than meetings — but check each listing: some roles are full-time.

Frequently asked questions

Can coders do AI training work part-time?

Often. Much of the work is async and queue-based, which fits around a primary job — but some listings are full-time or on-site roles, so check each one.

How much do AI training jobs for coders pay?

The 255 hourly listings here state a median top rate of $90/hr; the highest is $300/hr. Pay depends on credential depth, specialty area, and whether you pass into the higher-pay queue (most platforms gate this behind a more rigorous skills test than the entry queue).

What does AI training work actually look like for coders?

Reviewing AI-generated code for bugs, style, and correctness Writing reference solutions for prompts the model needs to learn from Grading multi-step agent runs that read/edit/run code Refactoring legacy code with the model as a pair, then scoring its suggestions

What's the catch?

The high-pay queue is gated behind a real coding test — usually a 60-90 minute screen with edge cases. Pay scales with language depth. Python + JS pay the entry-tier rates; deep C++, Rust, embedded, or low-level systems work pays more. Much of the coding evaluation work is queue-based and async rather than meetings — but check each listing: some roles are full-time.

Do I need to leave my current job?

Not for most of it: much of the work is queue-based and async. Some listings are full-time roles, and say so.

Other persona guides

See also: the pay-tier hub, scam detection, the 5-Q matcher.

Find your match

3,130 active AI training jobs. Listings last checked .