We are seeking a Technical Reviewer on behalf of a leading AI lab to evaluate and refine benchmarking pipelines for reinforcement learning (RL) environments and agentic AI systems. In this role, you’ll be responsible for reviewing environment design, terminal conditions, and evaluation protocols to ensure accuracy, reproducibility, and fairness in benchmarking. You’ll work closely with researchers and engineers to provide technical feedback that strengthens experimental rigor and system reliability. You’re a great fit if you: Have a background in reinforcement learning, computer science, or applied AI research . Are experienced with RL environments . Understand benchmarking methodologies, terminal conditions, and evaluation metrics for RL tasks. Are comfortable reading and reviewing codebases in Python (PyTorch/TensorFlow a plus). Have strong critical thinking skills and can provide structured technical feedback . Care deeply about experimental reproducibility, fairness, and standardization in agentic AI. Are detail-oriented and capable of reviewing both theoretical formulations and implementation details . Primary Goal of This Role To review, validate, and improve reinforcement learning environment benchmarking pipelines, ensuring that terminal conditions, evaluation metrics, and system behaviors are robust, reproducible, and aligned with agentic AI research goals. What You’ll Do Review RL environments and evaluate terminal conditions for correctness and consistency. Assess benchmarking pipelines for fairness, reproducibility, and alignment with research objectives. Provide structured technical feedback on code implementations and documentation. Collaborate with researchers to refine evaluation metrics and methodologies . Ensure reproducibility by validating results across different runs, seeds, and hardware setups . Document findings and recommend improvements for environment design and benchmarking standards . Why This Role Is Exciting You’ll directly influence the reliability of benchmarking in agentic AI research . You’ll work on cutting-edge RL environments that test the limits of intelligent agents. You’ll help establish standards for evaluation and reproducibility in a fast-moving field. You’ll collaborate with researchers shaping the future of agentic AI systems . Pay & Work Structure You’ll be classified as a full-time hourly contractor to Mercor. Paid weekly via Stripe Connect, based on hours logged. 40 hours/week commitment with flexible scheduling. Remote and flexible working style. We consider all qualified applicants without regard to legally protected characteristics and provide reasonable accommodations upon request.
We are hiring a Computational Physics Specialist to create terminal bench–style problems and tasks involving agents within the physics domain, as part of a hiring effort for its AI lab partner building advanced agent-based infrastructure. This role is ideal for individuals with a strong foundation in physics and a working knowledge of programming who are excited to design real-world simulations and problem-solving environments. You will be building terminal-based benchmarks where agents are tested on their ability to reason through and solve physics-based challenges using code, commands, and logic — all within a realistic terminal interface. You’re a great fit if you: Have a background in Physics, Applied Physics, or Engineering Physics. Are familiar with computational methods such as numerical integration, differential equations, Monte Carlo simulations, or physical modeling. Have coding experience in Python, MATLAB, C++, or similar languages, and are comfortable in a Linux terminal environment. Understand core physics concepts (mechanics, electromagnetism, thermodynamics, wave dynamics, etc.) and can translate them into structured problem-solving tasks. Are interested in AI agents and want to explore how they can interact with simulations, models, and physics solvers in terminal-based systems. Can design step-by-step challenges where quantitative reasoning, code-based exploration, and scientific accuracy are required. Primary Goal of This Role To design terminal bench–style tasks where autonomous agents are challenged to solve physics-based problems through code, simulations, and reasoning. These environments will test agent performance in areas like equation solving, unit conversion, modeling physical systems, and interpreting simulation outputs in a terminal context. What You’ll Do Build secure, sandboxed environments where agents interact with physics simulations or problem-solving tasks via the terminal. Design benchmarks that require agents to: Calculate physical quantities Run simple code-based simulations Solve real-world physics problems using appropriate equations and reasoning Collaborate with AI engineers to define task specifications, rewards, and difficulty levels. Create tasks that mimic real scientific computing workflows — e.g., modifying parameters, parsing output data, or debugging physical models. Develop tools for benchmark reproducibility, task evaluation, and automated resets. Scale challenges from basic mechanics problems to multi-step, interdisciplinary simulations. Why This Role Is Exciting Help shape a new class of agent benchmarks grounded in scientific reasoning. Apply physics knowledge in a creative and technically deep setting involving simulation, automation, and AI. Contribute to a high-autonomy team pushing the boundaries of agent performance in complex domains. Pay & Work Structure You’ll be classified as an hourly contractor to Mercor. Paid weekly via Stripe Connect, based on hours logged. Part-time (20 hrs/week) with fully remote, async flexibility — work from anywhere on your own schedule. We consider all qualified applicants without regard to legally protected characteristics and provide reasonable accommodations upon request.