AI Code-Agent Evaluator: Frontier DevOps Engineer

Obsidian; New York (NY); On-site; USD 455,000 - 647,000; Part time

Job summary

Obsidian is seeking contributors to join a frontier code agents project, evaluating and improving frontier AI coding models through structured assessments, focusing on infrastructure engineering workflows and model evaluation.

You'll review model-generated implementations across AWS/Azure/GCP, Kubernetes, CI/CD, observability, and automation, identifying bugs and edge cases while applying professional engineering judgment in realistic scenarios, with compensation per accepted task.

Qualifications

• 2+ years of professional DevOps, SRE, or Cloud Engineering experience.
• Experience with AWS, Azure, GCP, Kubernetes, Terraform, CI/CD pipelines, or observability tooling.
• Regular use of AI coding agents such as Cursor, Claude Code, Codex, Windsurf, Gemini CLI, or similar tools.

Responsibilities

• Use frontier AI coding agents to complete and evaluate complex infrastructure engineering tasks.
• Review model-generated implementations involving cloud platforms, Kubernetes, CI/CD systems, observability, and infrastructure automation.
• Identify bugs, edge cases, reliability issues, and failure modes.
• Compare outputs from multiple frontier models and assess their strengths and weaknesses.
• Apply professional engineering judgment to realistic infrastructure engineering scenarios.

Skills

DevOps

SRE

Cloud Engineering

AWS

Azure

GCP

Kubernetes

Terraform

CI/CD

Observability tools

AI coding agents

Tools

Kubernetes

Terraform

CI/CD pipelines

Observability tooling

Job description

Obsidian is seeking contributors to join a frontier code agents project, evaluating and improving frontier AI coding models through structured assessments, focusing on infrastructure engineering workflows and model evaluation.

You'll review model-generated implementations across AWS/Azure/GCP, Kubernetes, CI/CD, observability, and automation, identifying bugs and edge cases while applying professional engineering judgment in realistic scenarios, with compensation per accepted task.