Hey there! Thanks for being here! I am an assistant professor in the computer science and engineering department at the University of South Carolina (USC). My research involves advancing the state-of-the-art in AI/ML by developing novel algorithms and methods for solving some outstanding problems in:

- Sustainable AI [funded by NSF]: We have several ongoing projects on sustainable AI, including ML Inference Pipeline Adaptation, to find the right tradeoff between High Accuracy, Cost-Efficiency, and Sustainability. We also develop design strategies and patterns that developers can use to build sustainability into their system as an explicit goal.

- Autonomous Systems and Robotics [funded by NASA]: e.g. (i) Our work on Autonomy for space lander missions to the Ocean Worlds, such as Europa and Enceladus. Our project, RASPBERRY-SI, in collaboration with CMU, UArk, the University of York, as well as testbed providers: NASA JPL (physical testbed, called OWLAT) and NASA Ames (virtual testbed, called OceanWATERS). (ii) our recent project on developing a framework for designing Modular Autonomy on ROS (MARS).

- Causal Reasoning [funded by NSF and Google]: e.g., (i) Our work on causal reasoning for performance debugging and optimizations in computer systems (Distributed, Hardware-Software, ML Pipelines), in collaboration with Christian Kaestner (CMU) and Baishakhi Ray (Columbia). (ii) As a former visiting researcher at Google, I worked on causal representation learning, enhancing the explainability of learned representation in deep neural networks via causal features for large-scale systems that rely on such representations for automated decision-making.

- Sciences and Engineering [looking for funding!]: e.g., (i) Our collaborative work on causal learning for cancer research in collaboration with Phillip Buckhaults (USC's College of Pharmacy) and optimal vaccine promotion for COVID-19 in collaboration with Gregory Trevors (USC's College of Education). (ii) Our collaborative work on deep learning for symbolic mathematics in collaboration with Kallol Roy (University of Estonia).

I am particularly interested in the theoretical foundations of Causal Representation Learning, Adversarial ML, AutoML, and Transfer Learning. In addition to theory, I am excited about Sustainable AI, ML for Systems, and Systems for ML.

Before joining the faculty at the University of South Carolina, I was a postdoc in the School of Computer Science at Carnegie Mellon University (USA), working with Christian Kaestner, and, before that, I was a postdoc in the Department of Computing at Imperial College London (UK). I received my Ph.D. in Computer Science at Dublin City University (Ireland), and my advisor was Claus Pahl. I received an M.S. degree (research-based) in Systems Engineering from Amirkabir University of Technology (Iran) in 2006, and my advisor was Saeed Mansour. I also received a B.S. in Math & Computer Science from the Amirkabir University of Technology (Iran) in 2003.

Important Links

• Update: As of January 2023, I do not plan to hire new Ph.D. students until further notice and I am so sorry that I may not be able to answer prospective student emails asking for available positions at AISys. You are always welcome to apply for a master’s or Ph.D. degree at the department. If you are selected for such a graduate program in our department, you will have the opportunity to apply for an assistantship (teaching or research) to help cover your stipend and tuition, either at the department (teaching) or within specific laboratories directed by our faculties (research). Depending on common research interests, you will also have the opportunity to reach out to our faculties to find the best match for mentoring.

• With Vijay Chidambaram, Romain Jacob, Neeraja J. Yadwadkar, and Ivo Jimenez, we co-founded JSys—a new diamond open-access journal for the systems community. I also co-chair two areas at JSys: Computer Architecture (with Devashree Tripathy) and Configuration Management (with Tianyin Xu). Please consider submitting your research work to JSys!

• Want to work with us? Read about AISys Lab.

• Want to know about AISys and our collaborators? Here you can find an overview of AISys lab members and their research.

• Want to know more about our research? Please check out our Core Research Areas and Publications.

• Want to join our weekly reading group? Please read about our AISys Reading Group.

• Want to join our student-led robotic team at USC? Read about Gamecock Robotics.

• Want to know whether I am available for AI consultancy? Please check out our Consultancy Services.

News

Older News…

Current Projects

AdaptiveFlow (2022–)

In this collaborative project, we develop infrastructure, runtime methods, and tools to enable performance optimization of ML inference pipelines by runtime dataflow restructuring, resource adaptations, model switching, and configuration tuning of model server knobs, such as batch sizes. This project is led by the University of South Carolina (US) and in collaborations with Queens Marry (UK), Technical University of Darmstadt (Germany), Vrije Universiteit Amsterdam (Netherlands), and industrial partners, including Roblox (US). It is one of the projects in my group related to Design for Sustainability in Computing.

• Papers:
• IPA: Inference Pipeline Adaptation to Achieve High Accuracy and Cost-Efficiency, JSys
• Reconciling High Accuracy, Cost-Efficiency, and Low Latency of Inference Serving Systems, EuroMLSys

AI for Symbolic Math (2021–)

Mathematics is one of the most precious human heritage. To preserve mathematical knowledge even beyond the human era, in collaborations with the University of Tartu, we develop new methods to solve symbolic math tasks (differentiation, integration, PDE, linear algebra, etc.) with machine learning models such as neural networks. The key ideas behind our approach are to exploit the underlying structures (e.g., tree-like structures) behind symbolic math tasks and transform the symbolic math tasks into tasks such as language translations that deep learning models can do well.

• Papers:
• Pretrained Language Models are Symbolic Mathematics Solvers too!, Arxiv (2021)

NASA RASPBERRY-SI (2020–)

In this collaborative project (USC, CMU, York, UArk, NASA JPL/Ames/GRC), we develop technologies that enable learning-based autonomous planning and adaptation of space landers.

• Papers:
• CaRE: Finding Root Causes of Configuration Issues in Highly-Configurable Robots, RA-L -> iROS'23
• CURE: Simulation-Augmented Auto-Tuning in Robotics, T-RO

Causal AI: Structure Learning and Transfer Learning (2018–)

In this project, we are addressing the following important questions from theoretical and empirical perspectives: (i) How to learn reliable causal structures from data and how to use the learned causal structures for identifying causal invariances across environments? (ii) How to use the causal structure learning and counterfactual reasoning based on causal invariances for systems problems? In particular, we are extremely interested to apply these theoretical advancements for an explainable and guided performance debugging of highly-configurable systems.

• Papers:
• Learning LWF Chain Graphs: A Markov Blanket Discovery Approach, UAI 2020
• Transfer Learning for Performance Modeling of Configurable Systems: A Causal Analysis, AAAI 2019

ATHENA (2019–)

Despite achieving state-of-the-art performance across many domains, machine learning systems are highly vulnerable to subtle adversarial perturbations. In this project, we propose Athena, an extensible framework for building effective defenses to adversarial attacks against machine learning systems.

• Papers:
• ATHENA: A Framework based on Diverse Weak Defenses for Building Adversarial Defense, Arxiv (2020)

FlexiBO (2018–)

One of the key challenges in designing machine learning systems is to determine the right balance amongst several objectives, which also oftentimes are incommensurable and conflicting. For example, when designing deep neural networks (DNNs), one often has to trade-off between multiple objectives, such as accuracy, energy consumption, and inference time. We developed FlexiBO, a flexible Bayesian optimization method, to address this issue.

• Papers:
• FlexiBO: A Decoupled Cost-Aware Multi-Objective Optimization Approach for Deep Neural Networks, JAIR

BRASS MARS (2016–)

How do you build a software system that can function for a century without being touched by a human engineer? This is the herculean task being undertaken by the DARPA Building Resource Adaptive Software System’s (BRASS) program. We, at UofSC with several collborators at CMU, have been developing learning mechanisms to be integrated with quantitative planning to adapt/reconfigure robots to overcome environmental changes at runtime.

• Papers:
• Machine Learning Meets Quantitative Planning: Enabling Self-Adaptation in Autonomous Robots, SEAMS 2019
• Model-based Adaptation for Robotics Software, IEEE Software (2019)
• Transfer Learning for Improving Model Predictions in Highly Configurable Software, SEAMS 2017