Postdoc: Quantized RL & Discrete World Models

Freelio; Odense; Hybrid; DKK 440,000 - 580,000; Full time

Job summary

The University of Southern Denmark invites applications for a two-year postdoctoral research fellowship in quantized reinforcement learning, based in the ADIN Lab at IMADA, Odense. The project seeks to ground and implement embodied quantization, linking generative representations with discrete world models in non-stationary RL settings.

Applicants should have a PhD in a relevant field, a track record of first-author ML publications, and strong Python/PyTorch/JAX skills for developing scalable

Qualifications

• PhD in Computer Science, Mathematics, Statistics, or Theoretical Physics.
• Two first-author papers at flagship ML venues (NeurIPS/ICML/ICLR/AISTATS).
• Theoretical rigor in statistical learning theory and RL foundations, with analysis abilities.
• Strong Python, PyTorch/JAX programming and experience scaling deep models and RL pipelines.
• Experience with public code releases is a plus.
• Excellent written and spoken English communication.

Responsibilities

• Bridge high-capacity generative representations with discrete world models for RL.
• Develop uncertainty-aware world models and prove performance/ regret guarantees.
• Build scalable prototypes and benchmark across high-dimensional simulation suites.
• Publish in top-tier venues and fulfill teaching assistant duties.

Skills

Python

PyTorch/JAX

Reinforcement learning

Mathematical maturity

English communication

Education

PhD in Computer Science, Mathematics, Statistics, or Theoretical Physics

Tools

PyTorch

JAX

Git

Job description

The University of Southern Denmark invites applications for a two-year postdoctoral research fellowship in quantized reinforcement learning, based in the ADIN Lab at IMADA, Odense. The project seeks to ground and implement embodied quantization, linking generative representations with discrete world models in non-stationary RL settings.

Applicants should have a PhD in a relevant field, a track record of first-author ML publications, and strong Python/PyTorch/JAX skills for developing scalable