Position: Swarm Bench Task Engineer Data Analysis Type: Short-Term Contract (4 weeks) Compensation: $15 per hour Location: Remote Commitment: 30-40 hours per week with 4 hours overlap with PST
Role Responsibilities Design and author multi-agent benchmark tasks centered on complex data analysis workflows Create realistic synthetic datasets or curate real-world style datasets across domains such as finance, operations, security, or market analysis Build tasks that require agents to perform cross-referencing, anomaly detection, contradiction identification, and statistical computation across multiple sources Develop decomposition guides that split analytical work across specialist sub-agents such as financial, technical, security, or operations analysts Write precise oracle logic or verification scripts that validate specific analytical conclusions rather than generic summaries Create reproducible evaluation environments using Python and Docker Review task performance signals to ensure strong separation between weaker and stronger agentic systems Refine tasks to improve determinism, clarity, difficulty, and scoring quality
Requirements Strong years of experience in data analysis Strong proficiency in SQL and Python for data analysis and scripting (pandas, Num Py, or similar) Experience working with real-world, messy datasets such as CSV, JSON, logs, and reports Ability to design non-trivial analytical questions with clear, specific, and verifiable answers Solid understanding of statistical concepts including averages, distributions, outliers, and correlations Familiarity with AI coding benchmark environments (e.g., SWE-bench, Terminal-Bench) Comfortable with Docker including writing Dockerfiles, building images, and debugging container issues Ability to work independently in a remote environment
Application Process Apply/Easy Apply and check email for application form Fill Google form Assessment Link (After shortlisting to be completed within 24 hours)

Skills: Python, Data Analysis, information technology