Scientific Coding Expert for Physics and Python AI Training
Listing checked September 8, 2026 · pay as published by Turing
Overview
Turing's SciCode project builds a rigorous STEM dataset designed to train and assess frontier AI models. As a SciCode Trainer, you'll write scientific coding challenges rooted in physics and use python to craft verified solutions and tests. Each task weaves one core problem with several connected sub-problems that grow in complexity. Your work directly shapes how AI handles complex reasoning in science and programming. It's a contract position requiring a 4-hour overlap with PST and a commitment of 40 hours weekly for 8 weeks.
What You'll Do7
- 1Develop scientific problem sets, each with one central challenge and at least three sub-problems that build logically toward a complete solution.
- 2Create verified reference solutions in python that pass all unit tests.
- 3Design test cases that tell correct and incorrect model answers apart with high precision.
- 4Conduct QC checks on the Turing Central Task Platform, covering structural and quality standards.
- 5Revise tasks based on QC feedback to meet Pass@K benchmarks across LLM judges such as GPT, Gemini, and Nemotron.
- 6Consistently produce top-tier output on the first attempt, minimizing the need for rework.
- 7Attend sync calls for project reviews, feedback, and standups during the required coverage window.
Requirements7
- 1A Master's or PhD in physics or a closely related field.
- 2Strong python coding skills with hands-on experience in scientific computing.
- 3Capability to frame well-posed scientific challenges with clear constraints and expected outputs.
- 4Meticulous attention to detail to satisfy strict quality rubrics for correctness, determinism, and test sensitivity.
- 5Background in AI data annotation, research, or scientific writing, plus comfort with LLM evaluation frameworks or coding benchmarks.
- 6Experience using numpy, scipy, sympy, or other specialized scientific libraries.
- 7A record of published research or academic projects in a STEM area.
Who Should Apply
This role fits a physicist who codes with confidence and enjoys breaking down complex problems into testable pieces. If you have authored research or built scientific tools, you'll find the work rewarding. It's less suited for someone without deep physics knowledge or who struggles with strict quality guidelines. Rejection often happens when candidates underestimate the precision needed for deterministic test cases or can't commit to the 4-hour PST overlap. Another common pitfall is overlooking the need for continuous output consistency over the 8-week contract.
Pay and demand for Machine Learning & AI roles
AggregatedTypical pay
$70/hour
This role
Pay not disclosed
Most Machine Learning & AI roles pay $50–$100 per hour.
Based on 607 similar roles that publish pay · 102 publish only a top rate; those count at the rate they gave
Rates shown per hour. Yearly and monthly pay converted; one-time fees and non-USD pay are not included.
- Live similar roles
- 692
- Listed in last 30 days
- 306
- Remote
- 97%
Hiring most right now: micro1 (358) · SME Careers (111) · Mercor (104)
Most requested skills · share of roles
- python14%
- llm evaluation8%
- ai training6%
- quality assurance6%
Figures from Machine Learning & AI roles live on NearSkill when this page loaded. A role can close before you apply, so check the listing itself.
Compare your resume against these rolesLocation
Compensation
Undisclosed by Turing
Comparable roles pay
607 roles$50 – $100/ hr
Not this role’s pay. Turing has not published a range.
Confirmed during Turing profile screening.
Required Skills
Application Tip
Showcase a reproducible example of a scientific coding task you've designed, highlighting how you structure sub-problems and validate test cases.
See NearSkill jobs more often in your search
Application & verification flow
1Instant rubric match
Your resume is scanned against this role’s requirements to check qualification fit.
2Screened before the employer sees it
Only profiles that clear screening are passed on.
3Outcome by email
We notify you at the address on your resume once the screening is reviewed.
Similar open positions
Explore active roles that match your skills and interests.
Turing
VerifiedMaterials Science Coding Expert (Python)
Turing’s SciCode project is assembling one of the most demanding STEM AI training datasets in the industry. As a SciCode Trainer, you will craft, code, and validate intricate scientific problems in materials science that push frontier AI models to their limits. This direct contribution supports AI research by challenging models with precise, discriminative tasks. You will verify solutions in Python against strict quality rubrics and iterate with feedback from multiple LLM judges.
Turing
VerifiedBiology Python Scientific Coding Contractor (Remote)
Turing, in partnership with NVIDIA, is assembling a rigorous STEM AI training dataset through the SciCode project. This initiative focuses on creating high-quality scientific coding tasks that help train and evaluate frontier AI models. As a SciCode Trainer, you will craft, implement, and review intricate scientific problems in biology, directly contributing to AI research that pushes the boundaries of what models can solve. Expect to work within a structured environment that demands precision, scientific accuracy, and a deep understanding of both biology and Python.
Turing
VerifiedMathematical Coding Specialist (Remote)
Turing is assembling one of the toughest STEM training datasets in AI, and the SciCode project sits at the center. You will design original scientific coding challenges that push frontier models like GPT, Gemini, and Nemotron to their limits. Each task you build becomes part of the training ground for the next wave of AI reasoning. Your work here directly influences how these models handle complex mathematics and real-world scientific computation. With a Python-heavy stack and a strong focus on rigor, the role suits someone who enjoys turning intricate math into clear, testable code.
Turing
VerifiedChemistry Python Coding Trainer (Remote SciCode)
Turing is assembling a team of scientific coding experts for its SciCode project, an effort to build a rigorous STEM dataset for training frontier AI models. As a SciCode Trainer, you will author and review chemistry problems in Python, creating tasks that test and improve model reasoning. The work involves a blend of scientific writing, coding, and quality control, all aimed at producing data that meets strict evaluation standards. This is a remote, 8-week contract role requiring a 4-hour overlap with PST hours and a commitment of 40 hours per week.
Mercor
VerifiedPhysics PhD Coding Experts
This six-week contract places you inside a benchmark build for leading AI labs. You will author original, executable physics problems that current frontier models cannot crack, sourcing material from published papers, open-source repositories, or your own designs. The workflow runs through a Git-based pull-request system with automated checks, so comfort with Git and Docker is essential. You'll earn $70 per hour while contributing around 20 hours each week, fully remote.
Turing
VerifiedPhysics Expert (PhD/Master's)
Turing, a San Francisco-based research accelerator, helps frontier AI labs and enterprises build advanced AI systems. In this remote role, you will design and solve challenging Physics problems to probe the limits of large language models like ChatGPT. You will produce step-by-step solutions, collaborate with LLM researchers on evaluation goals, and propose new benchmarks spanning early undergraduate to PhD-level topics. This contract work also introduces you to AI-assisted analysis, which can sharpen your own research methods.