The Faculty of Information Technology of Czech Technical University in Prague offers 9 postdoc positions as a part of the CTU Global Postdoc Fellowship. This attractive two-year fellowship program provides an opportunity for excellent researchers who have recently completed their PhD to continue their research career at CTU in Prague. Fellows receive a two-year fellowship and become members of a team led by a mentor. The deadline for applying is 31 August.
The fellowship aims primarily at external international scientists who are currently conducting research abroad. Applicants must have completed their PhD within the last seven years (i.e. 2015 or later). The fellowship aims at authors (co-authors) of two or more publications in a journal with IF.
Applicants are advised to contact mentors should they require more details on the position/topic.
Modern astronomy has undergone a true paradigmatic shift from hypothesisdriven science focused on investigation of a single class of objects to the datadriven research based on explorative analysis of petabyte-scaled surveys of the Universe. Current astronomical high-performance digital detectors in observatories generate petabytes of raw data per night. The data is panspectral, ranging from radio through visible light to X-ray and gamma-ray frequencies, New domains are emerging, such as particle astrophysics (neutrinos) and gravitational-wave astronomy.
Most of astronomical data are publicly available through sophisticated networks of federated interoperable data archives based on the same standards for data storage, query and transfer called Astronomical Virtual Observatory.
Requirements to pre-process, store, and analyze this big data pushed the current information technology to its true limits. High-throughput preprocessing algorithms based on massively parallel GPU platforms using workflow orchestration systems such as Dask or Spark for distributed processing are needed to reduce the amount of stored data to sustainable size. Advanced visualisation tools became a part of many astronomical projects and result in increasing amount of multimedia content in on-line volumes of major astronomical refereed journals.
Heterogeneity, multidimensionality, and sparsity of more and more complex astronomical datasets need special storage formats (e.g., Parquet, HDF5, ASDF) for rapid searching, filtering, and data mining. Astronomical analysis of big sky surveys has recently been done in distributed cloud environments called Science platforms (e.g., SciServer, Astro Data Lab) where interactive data mining and visualisation experiments are done through dedicated web GUI or in Jupyter Hub directly launched in data centers storing Big Data archives.
We are developing a special hierarchical semi-sparse cube architecture to store such data. Our aim is to facilitate processing of these data using cuttingedge technologies, such as Map-Reduce frameworks, GPU-accelerated farms, or HPC supercomputers. Our research group closely cooperates with teams representing large-scale astroinformatics projects, such as IVOA, LSST, or Heidelberg Institute for Theoretical Studies.
We are seeking an enthusiastic researcher with computer science background interested in astronomy to work on the design of the hierarchical cube architecture and processing pipelines scalable to petabyte volumes.
A candidate should have a solid background in algorithms and data structures, database and networking technologies, parallel and distributed algorithms, and should have programming skills in Python, C, and experience with big data frameworks (Map-Reduce, Spark, Dask…). Basic knowledge of cloud environment using Docker, JupyterHub, JupyterLab or Google Colab, and interest in astronomy is an asset.
Multi-agent path finding (MAPF) is a problem of navigating multiple agents to their individual goal positions so that agents do not collide. The problem has been intensively studied recently from the perspective of heuristic search. However, there are still open questions especially in compilation-based approaches to MAPF and in how to reflect properties of real environments such a continuity of space and time in formal models of MAPF. Tests of novel concepts on real robots in a laboratory is an optional part of this research.
Project page (MAPF) Project page (roboAGE)
The ideal candidate should actively publish in artificial intelligence venues (AAAI, IJCAI, …) and/or in specialized venues focused on planning, heuristic search (ICAPS, AAMAS, SoCS, …), or robotics (ICRA, IROS, …) and in related journals. Strong programming skills preferably in C/C++ are required.
Parameterized or multivariate analysis became in the last two decades a standard approach for (NP-) hard computational problems. Here, in contrast to classical complexity, the efficiency of algorithms is not measured only with respect to the length of the input instance, but also with respect to a specified parameter. The outcome is a more fine grained complexity landscape which allows for a better prediction of the complexity of particular instances encountered. The goal is to design algorithms with running time that is polynomial in the input size and exponential only in the designated parameter. Fundamentally, the degree of the polynomial is required to be independent of the value of the parameter. Parameterized complexity provides a mathematical framework to analyze these algorithms and also to show that for some problems and parameters there is (probably) no such algorithm.
Be familiar with several of the following areas: Parameterized complexity, integer linear programming, structural graph theory, game theory, computational social choice. Experience in optimization and in particular approximation algorithms or hardness of approximation is appreciated but not required. The candidate’s high scientific potential should be witnessed by publications in the area of algorithms or discrete math in proceedings of highly ranked international conferences and/or journals.
Cyber-physical systems (CPS) are (as defined by the NSF) engineered systems that are built from, and depend upon, the seamless integration of computation and physical components. Since computation is currently being integrated into the vast majority of technical systems, CPS are becoming highly relevant for the technical sciences as a whole. Many CPS interact with humans, and hence the safety of such systems is of utmost importance. The postdoc will contribute to ensuring the safety of such systems by taking part in research that designs algorithms for proving the correctness, and especially, safety, of formal models of cyber-physical systems.
The ideal candidate combines the ability of formally precise reasoning and algorithm design with the familiarity with technical aspects of cyber-physical (embedded) systems.
This project is focused on synthesis of models of artificial intelligence and machine learning to programmable hardware as FPGA devices. The goal of the project is to enhance frameworks such as qkeras, HLS4ML not only to limited accuracy, but also to other aspects of the design and mapping to the target platform. Tightly binding AI frameworks and logic synthesis algorithms (tools) gives the possibility of united optimization of AI models with respect to their implementation and vice versa. On the logic synthesis side, aspects as energy consumption, timing constraints, and reliability will be considered.
Have knowledge of modern algorithms from artificial intelligence, especially neural networks and deep learning, including frameworks. Python and C/C++ programming skills and experience with hardware design for programable devices as FPGA including VHDL/Verilog languages. Basic knowledge of logic synthesis and optimization algorithms.
Currently, there is a large class of domain-specific domain modeling languages. However, since the subject area tends to evolve over time, developers have to complicate the DSML as well. As a result, DSML loses its flexibility and usability. The use of various, highly specialized dialects of DSML, reflecting various aspects of the subject area, but consistent with each other, seems to be much more effective. The use of ontologies and corresponding formal mathematical models when creating them can be used to ensure the consistency of different dialects of DSML. This is possible, since the DSML is based on a domain model, which can be an ontology. The main goal of the research proposed is to propose new formal mathematical models and algorithmic mechanisms for describing and organizing coordinated changes to the ontology and related DSML without the need to re-create the entire structure of the DSML in a certain application domain.
With the dawn of quantum computing, many of contemporary algorithms our security depends on are becoming insufficient. As of today, new candidate algorithms for post-quantum security are being put under scrutiny, bringing many new challenges in cryptology and engineering fields. Requirements necessary for assuring security in the post-quantum era are even more complex, considering not only confidentiality, but also authentication, data integrity, availability and more.
Research in this topic will be focused on methods of cryptanalysis of state-of-the-art post-quantum schemes, side-channel attack countermeasures, and quantitative and qualitative evaluation of these countermeasures.
Ideal candidate should have strong background in cryptology, mathematics and cryptographic engineering. The candidate shall have thorough overview of implementation attacks and related knowledge of statistical methods. The candidate should also have an experience with FPGA design and/or programming of embedded systems.
State-of-the-art hardware components of cryptographic systems require quality true random number generators (TRNG). Reliable key generators that are based on physical unclonable functions (PUF) are also in demand. Such key generation is very desirable from a security point of view because the key generated in this way remains the “secret” of the hardware itself. It is also important to study the behavior of the proposed PUFs and TRNGs in terms of their long-term response stability (in case of TRNG, entropy). The aim is to design new PUF and TRNG architectures that are suitable for the purpose of generation of quality TRNG output that also guarantee stable key generation based on PUF responses. This also includes studying and understanding the behavior of these components at the statistical level as well as at the physical/technological level. The security of designed solutions needs to be tested for resistance to known attacks. Therefore, part of the research is the investigation of new possibilities of combining active and passive physical attacks with the usage of classical cryptanalysis (linear, differential, algebraic).
An ideal candidate should have defended a Ph.D. thesis in the field of hardware security. Knowledge in cryptographic key generation and storage in hardware, hardware side-channel attacks, and basic knowledge of TRNG and PUF is expected. Experience in implementation of TRNG and/or PUF circuits in FPGA or ASIC is an advantage.
Fixed-parameter tractability and approximation algorithms are nowadays standard tools for design of algorithms for hard problems in the area of Computational Social Choice. Surprisingly, kernelization, a prominent technique in FPT algorithmics, is not used as often to tackle social choice problems. Kernelization is a formalism of safe data reduction which we believe does have its place in all research disciplines dealing with large and complex datasets. The most recent approach is the so-called lossy kernelization which on the one hand cooperates with approximation algorithms (unlike kernelization which can only be pipelined with exact algorithms) and on the other hand, allows circumventing hardness results (in exchange for introducing a possible loss in the quality of the solution).
Be familiar with several of the following areas: Parameterized complexity, integer linear programming, structural graph theory, game theory, computational social choice. Experience in optimization and in particular approximation algorithms or hardness of approximation is appreciated but not required. The candidate’s high scientific potential should be witnessed by publications in the area of algorithms or discrete math in proceedings of highly ranked international conferences and/or journals.
To apply for this program, you need the following documents in English:
The application must reach CTU by 31 August 2021, and should be sent to research@fit.cvut.cz by email or by post to:
Office of Science and Research
Faculty of Information Technology
CTU in Prague
Thákurova 9
160 00 Prague 6, Czech Republic
Applications will be assessed by a committee based on documents sent by applicants. The mentor has a strong vote in the selection process. An (online) interview will be organized preferably within the first half of September.
The final decision of the committee will be communicated to applicants by 15 October 2021. Fellows are expected to start their job at CTU from January 2022.
The person responsible for the content of this page: Bc. Veronika Dvořáková
