A common perspective on modern computing
We study new computing paradigms through the principles and methods of computer science, with particular attention to the complexity introduced by parallelism and distribution.
Our work asks questions such as:
- How can applications exploit increasingly heterogeneous computing resources?
- How can computation and data be distributed efficiently across nodes, accelerators, clouds, HPC systems, edge systems, and geographically separated facilities?
- How can programming models, runtimes, and workflows reduce system complexity without sacrificing performance or control?
- How can large-scale systems remain reliable, secure, reproducible, and energy-efficient?
- How should software evolve as processor architectures and computing paradigms change?
This perspective connects our different research activities and provides a common methodological foundation as computing technologies continue to evolve.
Cross-disciplinary
Our research is grounded in computer science, but the problems we address are rarely confined to a single discipline.
Applications are important to us not only as use cases. They provide new questions, constraints, and opportunities that can lead to advances in programming models, algorithms, runtime systems, architectures, and computing infrastructures … and collaboration oppportunities with researchers in science, engineering, medicine, industry, and other application domains.
Research directions
High-performance and heterogeneous computing
We study programming models, runtime systems, performance engineering, and system software for multicore processors, accelerators, heterogeneous architectures, and large-scale HPC platforms.
Our interests range from abstractions for parallel programming to resource management, data movement, performance portability, energy efficiency, and hardware/software co-design.
AI at scale and generative AI
Artificial intelligence is increasingly a systems problem.
We investigate the parallel and distributed foundations of modern AI, including large-scale training and inference, generative AI workloads, distributed and federated learning, and the interaction between AI frameworks and computing infrastructures.
Our focus includes performance and scalability, but also communication, resource efficiency, reliability, privacy, security, data management, and the cost of moving computation and data across complex systems.
Scientific workflows and distributed computing
Modern scientific applications are often composed of many interacting computational and data-processing stages running across heterogeneous infrastructures.
We study workflow models and systems, orchestration, cloud-HPC convergence, cross-facility computing, data movement, storage, reproducibility, and the execution of complex applications across distributed resources.
Data analytics
Large-scale data analysis increasingly combines parallel computing, distributed storage, machine learning, and complex data-processing pipelines.
We study the systems and programming aspects of data analytics, with particular attention to scalability, performance, data locality, resource management, and integration with HPC and AI workloads.
Emerging processors and RISC-V
New processor architectures create opportunities to reconsider the relationship between applications, programming environments, runtime systems, compilers, and hardware.
RISC-V is an important part of this activity, together with accelerators and other emerging architectures. We investigate software ecosystems, vector processing, performance, energy efficiency, and hardware/software co-design.
Quantum and hybrid computing
Quantum computing introduces a different computational model, while practical quantum applications increasingly interact with classical, parallel, and distributed systems.
Our interest is particularly in this boundary: programming and execution models, hybrid quantum-classical workflows, simulation, orchestration, and the integration of quantum resources into larger computing infrastructures.