Masoud Daneshtalab, Professor

I am a Professor at Mälardalen University (MDH) and leading the Heterogeneous System research group (www.es.mdh.se/hero/). I joined KTH as European Marie Curie Fellow in 2014. Before that, I was a university lecturer and group leader at the University of Turku in Finland from 2012-2014. 

My research focuses on the theoretical foundations of centralized and distributed AI and deep learning algorithms, their practical applications in resource management, computer vision, biomedical fields, and algorithm-hardware co-design. My research vision encompasses four key areas (KA):

  • KA1: Robustness, Reliability, Fairness, and Security in AI
  • KA2: Generative AI and synthetic data
  • KA3: AI acceleration / AI algorithm-hardware co-design
  • KA4: Federated learning 


My group has a track record of developing methods and tools for optimizing AI/DL models using multi-objective neural architecture search (NAS), specialized pruning and quantization techniques and designing specialized AI/DL hardware accelerators. List of the open-source tools: https://www.es.mdh.se/hero/tools/

Summary of my leadership qualifications:

- Have (co-)led many research projects including: AutoFL, GreenDL, FASTER-AI, SafeAI, SafeDeep, AutoDeep, DeepMaker, DESTINE, PROVIDENT, HERO, AGENT, CUBRIC, ERoT, and µBrain with a total estimation of 160 MSEK; currently leading 6 AI projects (as PI).

- Have over 15+ years of teaching experience in computer science and AI in four countries: Uni. Tehran, Uni. Turku, Taltech, KTH and MDU (Sweden, Finland, Estonia, and Iran), and have developed more than 10 advanced- and 3 basic- courses in multiple countries and different programs (e.g. robotics, dependable systems, and applied AI).

- Have contributed to 2 international books, 8 book chapters, over 46 journal papers (10+ ACM/IEEE transaction journals) and over 200 reviewed international conference papers.

- Supervisor of over 9 passed PhDs and postdocs since 2011.

- Co-leading the heterogenous system research group (HERO) with 10+ PhD students and 3+ postdoc since 2018.

- Associate editors of journals of Elsevier MICPRO & MDPI Imaging

- Technical program committee of 20+ major conferences in AI and design automation

-  General chairman, vice-chairman, and steering committee member of multiple conferences.

-  Have been on the Euromicro board of directors and a member of the HiPEAC network since 2016.

- Have collaborated with more than 50 international institutes and co-authors with more than 221 scientists (according to DBLP record) 

5 grant awards for excellence in research from the Nokia Foundation, Kaute Foundation, Ulla Tuominen Foundation, Nanotechnology Initiative Council, and Telecommunication Research Center.

- Multiple evaluation committees: Belgian Research Council, Irish Research Council, European Horizon, Austrian Science Fund, Natural Sciences and Engineering Research Council of Canada.

 
  • Focusing on core AI principles via pioneering novel theoretical algorithms to address performance, reliability, security, robustness, and fairness concerns in AI models.
  • AI algorithm-hardware co-design / AI accelerator 

[Show all publications]

[Google Scholar author page]

Latest publications:

Bridging the Gap: An Interface Architecture for Integrating CAN and TSN Networks (Dec 2024)
Aldin Berisa, Saad Mubeen, Masoud Daneshtalab, Mohammad Ashjaei, Mikael Sjödin, Benjamin Kraljusic , Nejla Zahirovic
MRTC Report, Mälardalen Real-Time Research Centre (MRTC 2024)

Enhancing Drone Surveillance with NeRF: Real-World Applications and Simulated Environments (Oct 2024)
Joakim Lindén, Giovanni Burresi , Håkan Forsberg, Masoud Daneshtalab, Ingemar Söderquist
43rd Digital Avionics Systems Conference (DASC) (DASC'43)

Autonomous Realization of Safety-and Time-Critical Embedded Artificial Intelligence (Mar 2024)
Joakim Lindén, Andreas Ermedahl, Hans Salomonsson , Masoud Daneshtalab, Bjorn Forsberg , Paris Carbon
Design, Automation & Test in Europe Conference (DATE'24)

A Systematic Literature Review on Hardware Reliability Assessment Methods for Deep Neural Networks (Jan 2024)
Mohammad Ahmadilivani , Mahdi Taheri , Jaan Raik , Masoud Daneshtalab, Maksim Jenihhin
ACM Computing Surveys (CSUR)

Analysis and Improvement of Resilience for Long Short-Term Memory Neural Networks (Oct 2023)
Mohammad Ahmadilivani , Jaan Raik , Masoud Daneshtalab, Alar Kuusik
36th IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems (DFT 2023)

Efficient On-device Transfer Learning using Activation Memory Reduction (Sep 2023)
Amin Yoosefi , Seyedhamidreza Mousavi, Masoud Daneshtalab, Mehdi Kargahi
International Conference on Fog and Mobile Edge Computing (FMEC)

MSc theses supervised (or examined):
Thesis TitleStatus
OBJECT RECOGNITION THROUGH DEEP CONVOLUTIONAL LEARNING FOR FPGA finished