PostDoc in Multi-agent Reinforcement Learning for Robotic Construction
Your mission : The EPFL SYCAMORE laboratory lead by Prof. Maryam Kamgarpour and EPFL CRCL lead by Prof. Stefana Parascho are looking to hire a postdoctoral researcher at the intersection of reinforcement learning and multi-robot construction. You will collaborate with both of our labs on an innovative project, to design, verify and evaluate multi-agent reinforcement learning algorithms for multi-robot assembly tasks.
This position is funded by the Center for Intelligent Systems at EPFL.
EPFL launched the Center for Intelligent Systems (EPFL CIS) in October 2019. This center is a joint initiative of the schools ENAC, IC, SB, STI and SV that integrates research in Artificial Intelligence (AI), Machine Learning (ML) and Robotics to foster integrative research streams leading to the development of Intelligent Systems. Intelligent Systems perceive their environment, can learn from the data they collect and adapt to the changing world around them. Intelligent Systems can take many forms. When available, such systems will have profound implications for many areas including manufacturing, transportation, commerce, employment, healthcare, government, legal, security, privacy, and education. Four integrative research pillars have already been launched since the inception of CIS: “AI for Medicine”, “Intelligent Assistive Robotics”, “Decentralized edge AI Infrastructure” and “Digital Twin”. Furthermore, the EPFL Center for Intelligent Systems supports its 75 associated professors in their teaching and training mission of future engineers/scientists. In addition, the EPFL CIS serves as a “one-stop-shop” for questions from industry, society and politics on the topics of artificial intelligence, machine learning and robotics and pro-actively promotes technology transfer from research to society and industry by organizing thematic conferences such as the EPFL Digital Twin Days in November 2021 or thematic events such as the AI for engineering roundtable during the EPFL engineering industry day 2023. Finally, through its work and contacts, the CIS integrates EPFL into European and international research and excellence networks.
The project aims at increasing robots’ impact on a sustainable environment, by expanding their autonomy. Multi-robotic assembly applications have shown great potential for the efficient construction of structures in controlled environments but have yet to be employed in uncontrolled ones. The approach we pursue in increasing robot autonomy in construction is based on the development of multi-agent reinforcement learning theory and algorithms for autonomous multi-robot assembly and construction and their transfer to the physical world. Reinforcement learning has a great potential to find solutions to the highly complex problems of multi-robot assembly: sequencing, path-planning, task-allocation. However, the assembly application brings additional theoretical and algorithmic challenges for multi-agent reinforcement learning. These challenges include very sparse rewards capturing task completion, heterogenous state information of each robot due to a given robot using local sensors, and a very large state-space due to all possible configurations for the assembly task. In addition, design considerations are key to achieving not only a functioning construction process, but one that exceeds human design and construction capabilities. The outcome of the postdoc will result in provably convergent and efficient multiagent RL algorithms and their implementation and evaluation on a multi-robotic testbed, in collaboration with doctoral students of both labs.
Main duties and responsibilities include : Perform original research at the intersection of reinforcement learning and multi-robot assembly
Your profile : Applicants should have completed or be close to completing a PhD in reinforcement learning, with a focus on application of algorithms in robotics. A strong scientific background and a proven publication record in the above fields is required. Experience in construction robotics is a plus, as well as a documented interest in design. Excellent communication skills in English are expected.
Postdoctoral Position in Deep Learning for Structural Biology and Protein Design
Your mission : Project
The successful candidate will collaborate with both laboratories to develop and train deep learning models for modeling protein allostery and designing protein switches. Proteins are not static molecules. They often behave as switches, alternating between states that carry out distinct functions. Hence, switching is a powerful mechanism of biological regulation and typically occurs upon structural changes triggered by a wide variety of external stimuli (e.g. from photon absorption to the binding of another protein) through a process referred as allostery. Understanding how this switching behavior occurs at the molecular level remains challenging. The advent of deep learning offers new opportunities to explore and predict protein motions but these approaches have mostly been applied to static representations of protein structures. Here, the candidate will tackle the modeling of switching dynamics using specially designed geometric deep learning techniques applied to molecular representations. We expect that this approach will lead to a new understanding of the mechanisms underpinning these switches and, eventually, to generative approaches allowing to engineer novel molecular switches.
Our laboratories
http://barth-lab.epfl.ch: We belong to the Institute of Bioengineering at EPFL and are also part of the Ludwig Institute for Cancer Research (LICR) in Lausanne. While EPFL provides a world class environment for basic science and engineering, the LICR fosters translational applications of basic discoveries to cancer medicine at the highest level. Our laboratory benefits from this dual exciting environment and interdisciplinary approaches are essential to our research. We work at the interface of computational biology, biophysics, chemistry, and cell biology to uncover the molecular principles that regulate protein and cellular signaling. Using this understanding, we (1) model and design novel protein biosensors and protein signaling networks for synthetic biology and engineered cell therapeutic applications; (2) design enhanced and selective molecular therapeutics; (3) predict the effects of genetic variations on protein structure/function and protein signaling networks for personalized cancer medicine applications. We are part of the Rosetta Commons, a community of developers of the software Rosetta (https://www.rosettacommons.org), the premier suite for macromolecular modeling, and are actively developing novel computational approaches.
http://lts2.epfl.ch: We are with the School of Engineering and our research focuses on developing Machine Learning methodologies. The laboratory environment is extremely multi-disciplinary, hosting projects in AI for computational biology, computational chemistry or neuroscience as well as more theoretical work. Our main expertise is in geometric deep learning and in particular machine learning on graphs, but we recently investigated generalized Implicit neural representations and other representation learning architectures that can handle multi-view data or dynamics.
Your profile : PhD in Physics, Math, Chemistry or Computer Science. Ideal candidates will have strong programming skills in python, C/C++ python and demonstrated expertise in deep learning and macromolecular modeling. In addition, candidates should have a record of relevant publications in peer-reviewed international journals, the ability to speak and write effectively, strong team skills, be self-motivated, and creative.
We offer : EPFL provides state-of-the-art facilities and is one of the leading technical universities worldwide. A competitive salary is offered.
Interested applicants should upload the application online in one PDF file, which includes a curriculum vitae, a statement of research interests, and names of three references.
Start date : This position will open starting July 1, 2023.
Term of employment : Fixed-term (CDD)
Duration : 1 year CDD, renewable
Remark : Only candidates who applied through EPFL website or our partner Jobup’s website will be considered. Files sent by agencies without a mandate will not be taken into account.
