Job Type
Versátil Contract (Remote) Job Summary: We are hiring experienced reinforcement learning engineers and researchers, including research scientists, faculty, and experienced postdoctoral researchers, for a contract project that tests whether frontier AI models can carry out real reinforcement learning work. You will design research-grade tasks and environments that a leading AI model cannot solve, then review how and why the model fails. What you will do
• Design self-contained tasks drawn from the practice of reinforcement learning, such as implementing an algorithm like proximal policy optimization or soft actor-critic from a paper’s description alone, finding the bug in a flawed reward function or environment, reproducing a reported learning curve and explaining a discrepancy, implementing a custom advantage estimator or exploration strategy correctly, and running a sound statistical comparison between two agents across multiple random seeds.
• Build reinforcement learning environments in which AI agents complete multi-step work, including simulated environments, software and tool-use environments, browser and graphical interface environments, and command line environments, each with clear success criteria and reward signals that cannot be gamed.
• Build everything each task needs, including code, simulators, data, and dependencies, write a reference solution, and write automated tests or graders that verify whether a solution is correct. Because reinforcement learning results vary widely between runs and seeds, tests need to be deterministic or use well-justified tolerances.
• Run your task against a frontier AI model, review its attempt with the same rigor you would apply to a paper under peer review, and refine the task until the failure reflects a real gap in the model’s capability rather than ambiguity,



trick wording, or a reward the agent can exploit.
• Work with reviewers to bring each task to acceptance. Who we are looking for
• A PhD or Masters degree in machine learning, computer science, statistics, robotics, or a closely related field, or equivalent industry experience building and training reinforcement learning systems.
• Industry or academic experience in reinforcement learning research or engineering. Years spent on doctoral research count toward this total.
• At least one publication, ideally peer-reviewed work at a recognized machine learning venue or journal.
• Deep working knowledge of Python and at least one major framework, such as PyTorch or JAX, and hands-on experience with reinforcement learning libraries and environment interfaces such as Gymnasium.
• Fluency in the Linux command line, shell scripting, Docker, and Git.
• Experience as a peer reviewer, or in designing course assignments and automated grading, is a strong advantage, because the work draws on the same skills. Why Cobalt AI
• Advance frontier AI where it counts. Apply your research expertise to the data that frontier labs cannot obtain any other way, where your reasoning directly shapes how the next generation of models works through technical problems.
• Grow professionally. Expand your influence through evaluation projects, advisory roles, and research collaborations, while deepening your understanding of how frontier models are trained and assessed.
• Work with a top-tier network. Collaborate with researchers from leading institutions and labs on high-impact, versátil work.
• Set your own schedule. Flexible 10 to 40 hour weeks that fit around your research position and your life.
• Competitive pay. Rates vary by project and are determined by a number of factors, including scope, skillset, and experience.

📌 Machine Learning Engineer (México)
🏢 COBALT
📍 México