RL Intern: Build Autonomous Game Agents

Razer; Singapore; On-site; SGD 16,000 - 22,000; Part time

Job summary

Razer is seeking a Reinforcement Learning Intern to join the AI Software team in Singapore. The role focuses on developing autonomous agents that learn from virtual environments, with duties spanning environment integration, reward design, policy training, evaluation, and optimization for gameplay and automated testing.

You will design and evaluate RL agents using Gym or Unreal Engine, implement reward functions and training pipelines, and collaborate with mentors to interpret results, while

Qualifications

• Pursuing a Bachelor's or Master's in Computer Science, AI, or related field.
• Proficiency in Python and familiarity with machine learning fundamentals.
• Coursework or experience in reinforcement learning or simulation-based AI is advantageous.

Responsibilities

• Join our AI Software team as a Reinforcement Learning Intern and help develop autonomous agents capable of interacting with and learning from virtual environments.
• Design, train, and evaluate reinforcement learning agents using frameworks such as Gym or Unreal engine.
• Implement and test reward functions, policy optimization techniques, and training pipelines.
• Conduct experiments to measure agent performance and learning efficiency.
• Collaborate with mentors to refine models and interpret experimental data.
• Document processes, findings, and insights.

Skills

Python

Reinforcement Learning

ML Fundamentals

Team Collaboration

Education

Bachelor’s or Master’s student in CS/AI

Tools

Gym

Unreal Engine

Job description

Razer is seeking a Reinforcement Learning Intern to join the AI Software team in Singapore. The role focuses on developing autonomous agents that learn from virtual environments, with duties spanning environment integration, reward design, policy training, evaluation, and optimization for gameplay and automated testing.

You will design and evaluate RL agents using Gym or Unreal Engine, implement reward functions and training pipelines, and collaborate with mentors to interpret results, while