RL Env Engineer: Build Conversational Agent Training Envs

Besimple; San Mateo (CA); On-site; USD 140,000 - 200,000; Full time

Job summary

Besimple AI is building datasets, benchmarks, and training environments to help voice models understand people and complete real tasks. We seek an engineer experienced in building reinforcement learning environments to turn real-world interactions into trainable tasks for agents.

Your work will involve designing tasks, observations, actions, rewards, and termination conditions, as well as creating verifiers and scalable simulators for repeatable experiments in a fast-moving ML setting.

Qualifications

• Experience building or extending environments used to train or evaluate RL agents.
• Strong engineering skills and maintainable code.
• Experience defining task interfaces and debugging agent-environment interactions.
• Ability to analyze why an agent succeeds or fails and improve the environment accordingly.

Responsibilities

• Build training environments with defined observations, actions, rewards, and terminations.
• Design rewards, verifiers, and evaluation metrics for progress.
• Turn human demonstrations into tasks using real interactions.
• Develop scalable simulators and tooling to enable repeatable experiments.
• Collaborate with researchers to validate learning progress and transfer to new scenarios.

Skills

Reinforcement learning environments

Software engineering

Experiment design & evaluation

Agent-environment debugging

Education

Bachelor's degree in CS or related field

Tools

Gymnasium-style interfaces

RL training frameworks

Job description

Besimple AI is building datasets, benchmarks, and training environments to help voice models understand people and complete real tasks. We seek an engineer experienced in building reinforcement learning environments to turn real-world interactions into trainable tasks for agents.

Your work will involve designing tasks, observations, actions, rewards, and termination conditions, as well as creating verifiers and scalable simulators for repeatable experiments in a fast-moving ML setting.