OpenEnv Hackathon — Long-Horizon RL

Train an AI to recruit truck drivers through tool calls

A multi-turn RL environment where agents use CRM, messaging, approval, and workflow tools across 40-70 step episodes to screen candidates, avoid trap jobs, and close hires.

01

Tool calling

4 tools — CRM, messaging, approval, workflow. The agent must call the right tool with the right action at each step.

02

Long horizon

Episodes span 40-70 steps through a full recruiting pipeline: lead → contacted → interested → approval → offer → hired.

03

Hidden information

Driver preferences, deal breakers, and personality are hidden. Must be discovered through screening messages.

04

Trap jobs

Jobs that look perfect but violate deal breakers. Skip screening and you'll hire for the wrong one — big negative reward.

?

---

Stage
lead
Step
0 / 100
Reward
0.0
CRM
Messaging
Approval
Workflow
Discovered Info
No info yet — send messages and read replies
Reward
Steps
Final Stage