Skip to main content
StrideRecruit

AI Trainer (Physical Sciences)

micro1

Source: Himalayas — authorized feed

Location
Remote
Job type
Contract · Remote
Experience
Mid level
Posted
9 Nov 2026
Last seen live
10 Nov 2026
Source
Himalayas
Employer
micro1
View on Himalayas and apply ↗

This link opens the listing on Himalayas, where the application continues. It is not confirmed to be the employer's own website.

Job description

Pay: $60–$120/hour equivalent, paid per accepted task

Location: Global, fully remote

Job Type: Contractor

Time Commitment: 15+ hours per week

Schedule: Flexible

micro1 is looking for Computational Physical Scientists with prior professional AI-training or model-evaluation experience to support a project focused on improving advanced AI systems.

You will use your scientific expertise and programming skills to create challenging technical tasks, evaluate model-generated code and analyses, identify failures, and establish objective standards for correct solutions. This role is intended for candidates who have completed substantive AI-training work involving physical-science content.

What You’ll Work On

  1. Create realistic computational physical-science tasks that challenge advanced AI models.
  2. Develop task instructions, reproducible environments, reference solutions, and evaluation criteria.
  3. Evaluate model-generated calculations, code, simulations, models, and technical deliverables.
  4. Identify invalid assumptions, numerical errors, physically impossible results, incorrect methods, or reproducibility failures.

Required Qualifications

  1. Prior paid experience training, evaluating, benchmarking, or improving AI systems on scientific or technical tasks.
  2. Strong practical proficiency with Python.
  3. Deep expertise in at least one relevant physical-science specialization.
  4. Hands-on experience using computational methods for modeling, simulation, numerical analysis, data analysis, or reproducible research.

Relevant specializations may include physics, astronomy and cosmology, materials science, chemistry, computational physics, computational chemistry, and scientific simulation.

Public Proof of Skill

Strong applicants should provide evidence such as:

  1. Peer-reviewed publications or preprints involving computational methods
  2. Public simulation or analysis code
  3. Google Scholar, ORCID, arXiv, or equivalent research profiles
  4. Open-source scientific software contributions
  5. Reproducible notebooks, models, datasets, or simulations
  6. Public benchmarks, experiments, or technical projects

Preferred Qualifications

  1. Experience creating AI-training tasks, reference answers, rubrics, graders, or benchmark problems.
  2. Experience with NumPy, SciPy, SymPy, Astropy, PyTorch, JAX, or comparable scientific tools.
  3. Experience validating simulations or models against analytical results, observations, or experimental data.
  4. Availability to begin immediately.

Application Process

  1. Submit your application, AI-training experience, specialization, and public technical evidence.
  2. Complete an AI interview containing technical and integrity checks.
  3. Complete a hiring-manager review.
  4. If selected, complete project onboarding and begin your first assigned task.

Compensation

  1. Compensation is output-based. Experts are paid for tasks that meet project specifications and quality requirements. Task availability and completion time may vary.

Originally posted on Himalayas

Jobs provided by Himalayas · himalayas.app

About this job

This page describes one live job: AI Trainer (Physical Sciences) at micro1, listed as Remote.

Frequently asked questions

Where is this job based?
This job is listed as Remote.
What kind of work is this?
This is a Contract · Remote role at micro1.
When was this job posted?
This job was posted on 9 Nov 2026.
Where do I apply for this job?
Use the Apply link on this page. It opens the original listing on Himalayas, where you apply — StrideRecruit does not take applications.
Report this listing

Tell us if this listing looks wrong. We review reports; a report does not remove the listing by itself.

We do not ask for your name or email and do not store your IP address with the report.