Student Research Assistant Position in Generally Capable Autonomous Agents from Web-Scale Video Data

  • Stellenart:


  • Institut:

    Institut AIFB

  • Bewerbungsfrist:

    31. Juli 2024

  • Kontaktperson:

    Fechner, Marcus

  • Position: Student Research Assistant (Hiwi)
  • Department: Institute of Applied Informatics and Formal Description Methods
  • Supervisor: Marcus Fechner
  • Location: Karlsruhe Institute of Technology (KIT)


About Us

At the research group “Applied Technical-Cognitive Systems”, we are at the forefront of deep learning in the context of applied machine intelligence. Our research is in the area of autonomous systems, from self-driving cars (CoCar NextGen, CoCar and the shuttles Anna and Ella) to autonomous service robots. We utilize deep learning and other machine learning based approaches to advance these fields.

Position Overview
  • Position: Student Assistant
  • Start Date: As soon as possible
  • Working Hours: 20 - 80 hours per month
  • Duration: 6 months with possibility of extension


Job Description

Developing generally capable reinforcement learning agents poses a significant challenge, especially in hard exploration tasks. Expert robotic data is scarce and expensive, but also reward functions are not easy to design for complex tasks. On the other hand, unlabeled expert video data is abundant, but not straightforward to learn behavioral priors from, as no labels exist (actions).

In this research, we want to investigate how we can train agents on unlabeled expert videos, such as YouTube videos, in a scalable way to master a wide variety of complex tasks, not solvable by conventional reinforcement learning. For support on this topic, we are searching for a motivated student.

Your primary responsibilities will include:

  • Implementing and training deep learning models.
  • Reading related research papers and participating in discussions on the topic.
  • Collaborating with us on experimental design, execution, and evaluation.
  • Other tasks as assigned related to our research.


  • Current enrollment as a student at KIT.
  • Strong interest in deep learning and machine learning.
  • Knowledge of the programming language Python.
  • Knowledge of PyTorch and/or Tensorflow.
  • Experience in working with Linux and Git.
  • Motivated to read research papers.
  • Motivated, responsible, and a quick learner.
  • Speak either German and/or English.


What we Offer
  • Gain hands-on experience in the field of deep learning and conducting systematic research.
  • Work closely with experienced researchers and PhD candidates.
  • Weekly to bi-weekly meetings with supervisor.
  • Coding support and helpful supervision.
  • Flexible working hours to accommodate your class schedule and the option for working remotely.
  • Contribute to research projects and papers.
  • Access to top-of-the-line deep learning workstations with the latest GPUs.


How to Apply

If you are enthusiastic about deep learning and eager to contribute to our research, please send your application to marcus fechner∂kit edu with the following documents:

  1. Cover letter (0.25-0.5 pages): Why do you want to work on this topic? Why are you suitable for the position? (mention your interests and relevant skills).
  2. CV/Resume (max. 2 pages).
  3. Recent transcript of records / grading table.
  4. Optional: Any relevant coding or project portfolio.

If you have any questions or need further information, please contact marcus fechner∂kit edu. Natürlich auch gerne auf Deutsch :)