AI Technical Trainer Jobs

View detail

Conduct structured process interviews with leads across Finance, Legal, HR, Operations, and Business / Revenue to map recurring, high-cognitive-load tasks. Identify which tasks are suitable for Claude skill encoding — well-defined inputs, predictable structure, clear success criteria. Write, test, and iterate Claude skills: system prompts, instructions, behavioral rules, few-shot examples, output formats, and edge case handling. Validate each skill directly with end users until it performs reliably for non-technical team members. Maintain a documented skill library with scope, usage guide, known limitations, and update log for each skill. Build and share a reusable methodology for skill writing that other teams can apply independently after the internship.

View detail

Contributors may design original optics problems that simulate real physics research workflows, ensure these problems are computationally intensive and cannot be solved manually within reasonable timeframes, develop problems requiring non-trivial reasoning chains in mechanics, electromagnetism, thermodynamics, and quantum mechanics, base problems on real research challenges or practical applications from optics and physics practice, and document problem statements clearly with verified correct answers.

View detail

Contributors may design original optics problems that simulate real physics research workflows, ensure problems are computationally intensive and cannot be solved manually within reasonable timeframes, develop problems requiring non-trivial reasoning chains in mechanics, electromagnetism, thermodynamics, and quantum mechanics, base problems on real research challenges or practical applications from optics and physics practice, and document problem statements clearly and provide verified correct answers.

View detail

Design original optics problems that simulate real physics research workflows; ensure problems are computationally intensive and cannot be solved manually within reasonable timeframes; develop problems requiring non-trivial reasoning chains in mechanics, electromagnetism, thermodynamics, and quantum mechanics; base problems on real research challenges or practical applications from optics and physics practice; document problem statements clearly and provide verified correct answers.

View detail

Contributors may design original optics problems that simulate real physics research workflows, ensuring problems are computationally intensive and cannot be solved manually within reasonable timeframes (days/weeks). They develop problems requiring non-trivial reasoning chains in mechanics, electromagnetism, thermodynamics, and quantum mechanics, basing problems on real research challenges or practical applications from optics and physics practice. Contributors also document problem statements clearly and provide verified correct answers.

View detail

Contributors may design original optics problems that simulate real physics research workflows, ensure problems are computationally intensive and cannot be solved manually within reasonable timeframes, develop problems requiring non-trivial reasoning chains in mechanics, electromagnetism, thermodynamics, and quantum mechanics, base problems on real research challenges or practical applications from optics and physics practice, and document problem statements clearly with verified correct answers.

View detail

Design computational material science problems to challenge a frontier AI model using specialized tools such as ObsPy, instaseis, pyrocko, MITgcm, flopy/MODFLOW, or others. Each problem must have an answer verifiable by code and run inside a sealed Linux container with the tool pre-installed and a programmatic judge that grades the model's answer. Pick an anchor tool and design a problem hinging on its waveform-processing kernels, geophysical inversion routines, sub-surface flow solvers, or community-validated data pipelines. Write Python reference solutions, supply input files and model or domain definitions as needed. Decide numerical answers and acceptable tolerance for correctness. Test and tune the problem difficulty against batches of the AI model's attempts until the agent succeeds only infrequently. Submit the task for senior review to ensure quality. Calibrate problems by rewriting scenarios, tightening parameters, and observing AI model behaviors to achieve a pass rate of 10–30%. Learn and deepen command of the anchor tool and gain intuition for how the AI navigates complex scientific problems.

View detail

Design computational material science problems to challenge a frontier AI model requiring specialized tools. Pick an anchor tool and design a problem based on its waveform-processing kernels, geophysical inversion routines, sub-surface flow solvers, or data pipelines. Write a Python reference solution, supply input files and model or domain definitions where needed. Decide the numerical answer and the tolerance for correctness. Test the problem against the AI model in batches of parallel attempts, tuning difficulty until the agent succeeds in a small number of attempts. Submit tasks for senior reviewer feedback to ensure quality. Tune problems iteratively based on AI performance to achieve a 10–30% pass rate, rewriting scenarios and adjusting parameters as needed while gaining expertise in both the tool and AI behavior.

View detail

Design computational material science problems to challenge a frontier AI model that must have an answer verifiable by code and require a specialized tool like ObsPy, instaseis, pyrocko, MITgcm, flopy/MODFLOW, or others. Pick an anchor tool and design a problem based on its waveform-processing kernels, geophysical inversion routines, sub-surface flow solvers, or community-validated data pipelines. Write a Python reference solution, supply input files and model or domain definitions where needed. Decide the numerical answer and determine the domain-appropriate tolerance to count as correct. Test the problem against the model in batches of parallel attempts, tuning the difficulty until the agent only succeeds in a small number of attempts. Once finalized, the task is reviewed by a senior reviewer for quality feedback. Calibrate the problem by tuning it against batches of agent runs to reach a pass rate between 10–30%, involving rewriting waveform scenarios, tightening inversion parameters and solver tolerances, and monitoring agent behavior. Gain deeper command of the anchor tool and develop an intuition for how the frontier model navigates complex seismic, oceanographic, and sub-surface flow problems.

View detail

Design computational material science problems to challenge a frontier AI model, ensuring each problem has an answer verifiable by code and requires a specialized tool such as ObsPy, instaseis, pyrocko, MITgcm, flopy/MODFLOW, or others. Pick an anchor tool and design problems based on its waveform-processing kernels, geophysical inversion routines, sub-surface flow solvers, or community-validated data pipelines. Write Python reference solutions, supply necessary input files and model or domain definitions, decide on numerical answers with domain-appropriate tolerance, and test the problems against the AI model through batches of parallel attempts. Tune problem difficulty to achieve a pass rate in the 10–30% range, rewriting scenarios and adjusting parameters as needed. Submit problems for senior review to ensure quality standards. Gain deep understanding of anchor tools and AI model behavior through the iterative calibration and testing process.