To apply, email your CV to [email protected] with the subject "Terminal Bench Task Author". Full details below.

About the role
FMR Enterprises is recruiting senior infrastructure engineers for an AI evaluation project with a leading AI company. You will design Terminal Bench tasks. Each one is a realistic broken system that takes an experienced engineer hours to diagnose and fix, paired with an automated test that proves the fix works. Your tasks will be used to measure how well AI agents handle real production engineering.
What you will do
• Build broken Linux environments based on real production failures
• Write a reference solution showing how an expert finds the root cause
• Write automated verifiers that pass only when the system is truly fixed
• Refine your tasks based on reviewer feedback

Must have
• 3+ years in DevOps, SRE, platform or infrastructure engineering
• Strong Linux and terminal skills, fluent in Python and Bash
• Docker and Kubernetes, including debugging broken deployments
• Stateful systems: PostgreSQL, Redis,



message queues, replication and recovery
• Diagnosing ambiguous failures across multiple production services with no runbook
• CI/CD, infrastructure as code and cloud (AWS, GCP or Azure)
• A testing mindset: you write verifiers that cannot be gamed

Strong plus
• Authored tasks for Terminal Bench 2 or 3, SWE-bench Pro, DeepSWE, Senior SWE or similar benchmarks
• Nginx, systemd, LVM, security hardening, distributed systems
• Practical experience with AI coding agents

Details
• Fully remote
• Flexible schedule, hourly contract
• Short intro call after CV review

How to apply
Email your CV to [email protected] with the subject "Terminal Bench Task Author", or apply here.

To stand out, include:
• Links to any benchmark tasks you have authored (we review these first)
• Two or three lines on the hardest production failure you have debugged

📌 Terminal Bench Task Author | DevOps, SRE, Platform Engineering | Remote Contract (India)
🏢 FMR Enterprises
📍 India