Skip to content

Agentic & RL Envs

ML Engineer, LLM and Agentic Evaluations

Half of all Agentic Tasks & RL Environments roles come down within 9.3 days of being listed, and 34.1% are still up after four weeks. This one was posted today.

About the role

The work

  • Design, build, and run evaluation suites for reasoning, coding, mathematics, tool use, and long-context tasks.
  • Create benchmarks, task specifications, rubrics, and grading criteria for multi-step planning, function calling, code execution, and browser or computer-use agents.
  • Curate, clean, and annotate evaluation datasets while comparing open-source and frontier model outputs.
  • Build programmatic checks and model-based graders, then validate them against human reference judgments.
  • Investigate failure cases, identify error patterns, and write reproducible reports with practical recommendations.
  • Integrate evaluations into harnesses, CI pipelines, and dashboards.
  • Improve documentation and evaluation tools.

What it takes

This is a remote contractor opportunity for qualified contributors. The engagement is planned for 3 months.

You should be pursuing or have recently completed a bachelor's, master's, or PhD in computer science, machine learning, data science, mathematics, or a related quantitative discipline.

  • Open to candidates in India, Pakistan, and Turkey.
  • Part-time engagement requiring 20+ hours per week.
  • Fluent English.
  • Strong Python skills with familiarity with NumPy, Pandas, Jupyter, and standard machine learning workflows.
  • Understanding of how modern LLMs are trained, prompted, and evaluated.
  • Experience using LLM APIs or open-source models through tools such as Hugging Face or vLLM.
  • Strong analytical, technical writing, and independent problem-solving skills.
  • Ability to collaborate remotely with overlapping working hours.
  • Experience designing model benchmarks, evaluating agentic workflows, building automated graders, or analyzing human reference judgments is helpful.

How it works

Apply on OpenTrain with your resume, then complete the application on the hiring site.

About AI training work

AI training work uses human-created examples, evaluations, and feedback to improve how artificial intelligence systems behave. People with technical experience are needed to design reliable tests, review model performance, and identify where systems fail.

Where you can work

This role is 100% remote, for applicants located in India, Pakistan and Türkiye.

From OpenTrain AI's job posting, checked 18 Nov 2026.

About OpenTrain AI

A unified board of AI-training jobs with real per-role pay.

How it pays: Stripe payouts to your bank, in the 120 countries its payout docs list.

When it pays: When the employer approves each funded milestone. How often others pay

Interview: on some roles or projects (AI interview (switched on per job by the employer)). How the OpenTrain AI AI interview works

Similar open roles

Some links on AITraining.jobs are referral links. If you sign up through one, AITraining.jobs may earn a commission at no extra cost to you. It never changes what a platform pays you. A role at a platform that pays us a referral fee weighs more in this order.

An experienced engineer?

One profile for every AI training platform. Companies ask us for people with a specific language, country or credential: add yours once, and when a role fits, we introduce you, only with your yes. Free.

Create my profile

Get new OpenTrain AI roles by email

We check the OpenTrain AI job board every hour and email you its newest roles each week, plus similar paid roles elsewhere. Free, unsubscribe anytime.