Research Engineer, AI-based Communication Learning
This position involves working on a project related to communication learning for joint cooperation in large multi-agent systems. The objective of this position includes contributing to the model development, solution, and implementation of novel communication learning methods using advanced reinforcement learning tools such as Graph Neural Networks. These models, once trained, will also be validated and tested in simulation on standard benchmarks (e.g., Starcraft Multi-Agent Challenge), as well as on actual robotic hardware on a number of cooperative multi-robot tasks (e.g., multi-agent pathfinding, search, or collaborative manipulation). The role involves the opportunity to work with the research team within the Mechanical Engineering department at the National University of Singapore. The successful candidate will be self-motivated with an outstanding track record in computer science, robotics, machine learning (in particular, reinforcement learning), or related disciplines.
The main research tasks for the project include but will not be limited to:
- Developing novel decentralized communication learning methods to enable and improve cooperation in large multi-agent systems, based on local sensing and decision-making, and by relying on/extending recent reinforcement learning tools.
- Testing the devised models against state-of-the-art methods in standard simulation-based benchmarks widely used by the community.
- Implementing and testing the devised models on actual hardware, to experimentally validate their use and test their limit under real-life conditions.
- Coordinating a small team of researchers at different levels (e.g., FYP students, remote/on-site interns, master's students).
- Write papers and present work at top-tier local or international conferences or workshops.
- BEng/BSc/MEng/MSc in Computer Science, Engineering, Artificial Intelligence, Machine Learning, or related disciplines.
- Experience in developing communication learning models using recent reinforcement learning methods, and applying these models to standard benchmark tasks (e.g., Starcraft Multi-Agent Challenge).
- Interest and/or experience implementing machine learning models on robotic hardware.
- Proficiency with GNU/Linux, Python/C++ and at least one of the main machine learning libraries (e.g., tensorflow, keras, pytorch, CAFFE).
- Excellent interpersonal and communication skills.
- Good teamwork abilities, self-motivation and self-reflection.
- Proficiency in English language, both in oral and written form.
Location: Kent Ridge Campus
Department: Mechanical Engineering
Employee Referral Eligible: No