About Turing

Turing is one of the world’s leading AGI infrastructure companies, working with frontier AI labs to accelerate model development through high-quality training data, evaluations, and engineering talent.

Engagement Details

Compensation: Market, please provide a specific hourly rate expectationAvailability: 40 hours/week, with at least 6 hours of IST/dayType: Independent contractorDuration: Approximately 2 monthsStart: 01-Nov-2026

About the Role

We are looking for an engineer to build and deploy an intelligent, autonomous operations agent that supervises our data pipeline infrastructure around the clock. The agent will predict delivery delays before they occur, verify data quality using adaptive statistical baselines, and synthesize code-level fixes for routine operational failures. You will own the agent end to end, from design and evaluation to production reliability.

What You Will Do

Predictive health and SLA forecasting: Build models that analyze in-flight pipeline progress, execution velocity, and resource consumption against historical benchmarks to forecast SLA breaches 1-2 hours ahead, triggering priority re-queueing or upstream escalation.Bottleneck profiling and self-healing: Enable the agent to inspect execution DAGs and query profiles when jobs stall or degrade, isolate root causes, and autonomously create isolated branches that synthesize fixes for syntax errors, upstream schema changes, and column renames. Each fix is verified through dry-run assertions and submitted as a review-ready change request with full test results for human approval.Statistical and semantic data quality forensics: Replace static thresholds with rolling time-series baselines that account for seasonality, day-of-week swings, and business close cycles, and profile categorical entropy, column distributions, and foreign key orphan rates to catch subtle corruption that row counts and schemas miss.Lineage reconciliation and quarantine: Trace datasets from raw ingestion through marts and reporting views, run cross-tier checksums and consistency assertions before reporting cycles begin, and automatically quarantine partitions with severe violations so bad metrics never reach executive dashboards.Conversational copilot and incident management: Generate root-cause incident briefs covering failure cause, affected downstream assets, and recovery steps to cut alert noise, and build a natural language interface for checking pipeline health, triggering selective partition re-runs, and analyzing data distributions.