RL Environment Engineer — Build Conversational AI Tasks

BeSimple AI; San Mateo (CA); On-site; USD 140,000 - 230,000; Full time

Job summary

Besimple AI is building RL-powered training systems for voice agents. We seek an engineer with hands-on experience building reinforcement learning environments to turn real-world interactions into task-based simulations, observations, actions, state transitions, rewards, and termination criteria.

You will design rewards and verifiers, convert human conversations into training tasks, and develop scalable simulators that run reliably in training pipelines.

Qualifications

• Hands-on experience building or extending RL training environments.
• Strong engineering skills and tested, maintainable software.
• Experience defining task interfaces and environment state.
• Ability to investigate why agent fails or succeeds for the wrong reason.
• Clear technical communication and ability to take open-ended problems to a reliable system.

Responsibilities

• Build training environments with tasks, observations, actions, rewards, and termination conditions.
• Design rewards and verifiers to detect task completion and progress.
• Turn human conversations into training tasks from demonstrations.
• Build interactive simulators that respond to an agent’s actions and maintain state.
• Make environments scalable with rollout interfaces, logging, and debugging tools.
• Validate learning progress with baseline policies and experiments.

Skills

RL environments

Software engineering

Task interfaces

Debugging agent–environment

Job description

Besimple AI is building RL-powered training systems for voice agents. We seek an engineer with hands-on experience building reinforcement learning environments to turn real-world interactions into task-based simulations, observations, actions, state transitions, rewards, and termination criteria.

You will design rewards and verifiers, convert human conversations into training tasks, and develop scalable simulators that run reliably in training pipelines.