Columbia, South Carolina, United States
CS Prof. at UofSC (US); former visiting researcher at Google (US), ex-postdoc at CMU (US), ex-postdoc at Imperial College London (UK), Ph.D. from DCU (Ireland), B.Sc. (Math & CS) and M.Eng. (Systems Engineering) from Amirkabir University of Technology (Iran). Specialties: ML for Systems, Systems for ML, Causal AI (Theory and Applications), Autonomous Systems, Robotics
My goal is to advance a scientific and principled understanding of learning-enabled autonomous systems (e.g., rovers, spacecraft landers) informed by careful empirical and theoretical work. My research is driven by a theoretical understanding of machine learning and the mathematics behind statistical learning theory. I am, in particular, interested in transfer and representation learning, with the goal of developing reliable and robust algorithms that enable autonomous agents to learn optimal policies in a way that they can reuse the learned policy to solve similar tasks over varieties of diverse environments. Naturally, I am also interested in connections between computer systems, machine learning, and artificial intelligence. Details: https://pooyanjamshidi.github.io/AISys/
I worked on (Deep) Representation Learning, Causal Learning, and Neural Network Pruning and Quantization. I was also partially involved in testing our ATHENA framework (the main work has been done by my former Ph.D. student, Ying Meng). Please check it out: https://softsys4ai.github.io/athena/
Worked as a postdoctoral researcher in Institute for Software Research, School of Computer Science at CMU. My research at CMU concerns building machine learning methods for the analysis of robotics software with respect to various quality attributes such as performance, the automated understanding of properties and the runtime adaptation and repair of robotics software systems based on runtime performance reasoning. More specifically, my research will be about sensitivity analysis to understand the influence of changes in large configuration spaces (learning, sampling), the analysis of interactions among various changes (static or dynamic analysis, instrumentation), the integration of machine learning (transfer learning) to facilitate fast and reliable model learning for runtime analyses.
I'm involved in two European projects, FP7 MODAClouds (multi-cloud modeling and runtime adaptation) and H2020 DICE (DevOps, performance modeling, testing and architecture refactoring for big data systems). My current research is predominantly in the area of the usage of machine learning for enabling configuration optimization, self-adaptation and anomaly detection in distributed processing platforms which are designed to process big data and provide predictive analytics. I generally undertake research in self-adaptive software, cloud computing, performance engineering, big data and software architecture refactoring.
I undertake research on Self-Adaptive Software, Cloud Computing, IoT, Robotics
I undertake research on robust control of uncertainty in autonomic resource provisioning for cloud-based software.