Headquarters:
Overview
A fast-growing AI evaluation company is building expert-level benchmarks that test how well AI agents handle real professional work in the terminal. We are hiring cryptography experts to turn real, paid tasks from their own work into end-to-end challenges for AI agents, each with a working solution and a check that proves the result is correct.
***THIS IS A TASK BASED COMPENSATION MODEL. PLEASE CHECK THE APPLICATION QUESTIONS FOR MORE DETAILS.***
Responsibilities
- Design expert-level Terminal-Bench cryptography tasks by turning real, paid work from your own experience into challenges that an AI agent must solve end to end.
- Define the task environment, including files, systems, ciphertexts and data, so the challenge is realistic, self-contained and reproducible in a Linux command-line setup.
- Write a working reference solution in Python or SageMath that proves the task is solvable, for example by recovering a key or decrypting a secret.
- Build automated checks and tests that objectively verify the final result is correct, and explain how a mistake or a partial answer would be detected.
- Calibrate task difficulty so a typical expert needs about 60-75 minutes to solve it, keeping it hard for AI agents while still fair and unambiguous.
- Package each task with Docker and Git where applicable, document instructions clearly, and revise based on reviewer feedback until it meets the required quality bar.
Skills/Experience
- 3+ years of hands-on cryptanalysis or applied cryptography; GRC, audit-only, policy, sales or management roles do not count, nor does coursework only
- Did the work in the last 3 years and can name the systems, tools, data and rules used
