Skip to content

Tutorials

New to ParleML?

Start with Get started. It takes you from nothing to a trained, explained model in about 30 minutes, and every tutorial here assumes you've seen it.

Each tutorial follows one complete piece of work on an included sample, with screenshots from a recorded run. The recorded numbers are what that run produced; yours can differ slightly, especially with an agent.

Tutorial What you'll learn Time How
Customer churn with an agent Give an agent a bounded task, answer its question, and inspect one baseline and its drivers 30 min Agent
Build a claims review queue Steer an agent from a business question to a review queue: text fields, three decisions, clusters, rules, a local scoring demo and a staffing trade-off 2 h Agent
Steer an agent through a segmentation Put a stop rule in the brief, see the agent re-check your answer, and challenge its method with a follow-up 45 min Agent
Forecast retail sales by hand Check forecastability, race statistical models, add promotions as a driver, explain why it wins, register it 45 min By hand
Next best product for telecom customers Judge a recommender and a ranker against popularity, catch a weak sub-model, trade accuracy for coverage 30 min By hand
Segment customers and review unusual payments Cluster with automatic k, check against held-out personas, rank payments with three detectors, pick a threshold 30 min By hand
Take a forecast from review to production An agent challenges the model and asks for approval with conditions; you answer and move it to Production 30 min Agent review, your approval
Try a local agent on a laptop Connect Codex or Claude Code to a small Ollama model, check it end to end, and see its limits 20 min Agent (local model)

The agent tutorials need an agent signed in (Run ParleML, step 3). The by-hand tutorials need only ParleML running.

Details: which tutorials build on each other

For one job at a time, such as single sign-on or scoring a new file, see the how-to guides. For component examples, see the ML4 examples, the MLI examples and the agent cookbook.

Next