AI/ML Engineer - Time-Series

About the role

stepscale AI’s value comes from the models that learn workload patterns and recommend scaling configurations. You will own the modeling stack: feature engineering on noisy real-world metrics, training pipelines, deployment, and the evaluation framework that proves a recommendation is actually better than what the customer had.

What you will work on

• Forecasting models for queue depth, task utilisation, and processing rates across hourly / daily / weekly cycles
• Anomaly detection that separates predictable peaks from incidents
• The evaluation harness - synthetic + historical replays comparing AI-tuned configs against the customer’s current configuration
• The recommendation policy that turns a forecast into a concrete config diff (thresholds, min/max bounds, cooldowns)

What we are looking for

• 4+ years building production ML systems on time-series data - forecasting, anomaly detection, or similar
• Strong Python and a solid statistical foundation; deep-learning experience nice but not required
• Comfortable defending modeling choices with numbers, not vibes
• Bonus: experience with observability data (Prometheus, CloudWatch) or scheduler / autoscaling problems

How we work

Remote, async by default. You will be one of the first ML hires; expect to build the platform you and future engineers depend on. We give research time, but the bar is shipping models that customers run against.