Hi there!
I'm currently a Postdoctoral Researcher at IBM Research Zurich. You can read more about my ongoing research here!
Domain-Specific Training: Der heilige Gral für ChatGPT & Co
Abseits von “Spielereien”: Um LLMs (ChatGPT & Co) produktiv verwenden zu können, müssen sie verlässliche Antworten für definierte Anwendungen liefern können. Dieser Gretchenfrage und ihre zugrundeliegenden Herausforderungen sind Christian R. Ulbrich und ich in einem Gastbeitrag für Swiss Engineering STV UTS ATS nachgegangen. Wir diskutieren auch ob “fine-tuning” (also “gezielte Fortbildungen” für LLMs) eine Lösung sein kann… Der Artikel wurde Ende letzen Jahres hinter einer Paywall veröffentlicht, aber wir durften das PDF hier veröffentlichen.
Posted
#research #internet politics
Don't forget the compiler: Why FPGAs for HPC need to look beyond circuits and applications
FPGAs promise to accelerate HPC workloads and ML/AI models while also being energy efficient. However, today’s FPGA tool chains are cumbersome to use, limited to specific use cases and devices, and mostly fail to support workflows requiring multi-node application scenarios. Last month, I was invited to give a talk at HiPEAC’s F4HD workshop to deliver the message “Don’t forget the compiler: Why FPGAs for HPC need to look beyond circuits and applications”….
Posted
#research
Operation Set Architectures for organic FPGA ecosystems
How can we bring different optimized-down-to-the-gate FPGA libraries and latest state-of-the-art ML and AI together? In my research of 2022 and 2023, I answered this by inventing the Operation Set Architectures. The basic principle is to define operation blocks at an intermediate representation that is 1) low enough to enable compiler optimizations but 2) high enough to still allow easy architecture-specific optimizations. My research was published in the IEEE Compture Architecture Letters with Advancing Compilation of DNNs for FPGAs using Operation Set Architectures and the top-tier IEEE EDGE conference with DOSA: Organic Compilation for Neural Network Inference on Distributed FPGAs.
Posted
#research
Logikchips für die Hyperscaler-Cloud
Die Schweizer Zeitschrift Computerworld hat über unsere Arbeit für das IBM cloudFPGA projekt berichtet: Logikchips für die Hyperscaler-Cloud. Ein gute zusammenfassender Bericht über cloudFPGA auf deutsch!
Posted
#research
A Case for Function-as-a-Service with Disaggregated FPGAs
What does it take to realize highly dynamic cloud services like Function-as-a-Service with FPGAs? Find out in my recent paper titled A Case for Function-as-a-Service with Disaggregated FPGAs (DOI: 10.1109/CLOUD53861.2021.00047). Published in the Proceedings of the 2021 IEEE 14th International Conference on Cloud Computing (CLOUD). Abstract The slowdown of Moore’s law and the end of Dennard scaling created a demand for specialized accelerators, including Field Programmable Gate Arrays (FPGAs), in cloud data centers.
Posted
#research
Programming Reconfigurable Heterogeneous Computing Clusters Using MPI With Transpilation
How can the execution of a collective program optimized for heterogeneous CPU + FPGA clusters? A part of the answer is transpilation and you can find out more in my recent paper about Programming Reconfigurable Heterogeneous Computing Clusters Using MPI With Transpilation (DOI: 10.1109/H2RC51942.2020.00006). Published in the Proceedings of the Sixth International Workshop on Heterogeneous High-performance Reconfigurable Computing (H2RC’20). Abstract With the slowdown of Moore’s law and the stop of Dennard scaling, energy efficiency of compute hardware translates to compute power.
Posted
#research
ZRLMPI: A Unified Programming Model for Reconfigurable Heterogeneous Computing Clusters
How can a unified programming model for heterogeneous CPU + FPGA clusters look like? Find out in my recent paper about ZRLMPI: A Unified Programming Model for Reconfigurable Heterogeneous Computing Clusters (DOI: 10.1109/FCCM48280.2020.00051). Published in the Proceedings of the 2020 IEEE 28th Annual International Symposium on Field-Programmable Custom Computing Machines (FCCM). Abstract Over the past two decades, the Message Passing Interface (MPI) has evolved as the de-facto standard for programming High-Performance Computing (HPC) clusters.
Posted
#research
System Architecture for Network-Attached FPGAs in the Cloud using Partial Reconfiguration
How to configure and manage zillions of Cloud FPGAs? Find out in my recent paper about System Architecture for Network-Attached FPGAs in the Cloud using Partial Reconfiguration (DOI: 10.1109/fpl.2019.00054). Published in the Proceedings of the 29th International Conference on Field Programmable Logic and Applications (FPL). Abstract Emerging applications such as deep neural networks, bioinformatics or video encoding impose a high computing pressure on the Cloud. Reconfigurable technologies like Field-Programmable Gate Arrays (FPGAs) can handle such compute-intensive workloads in an efficient and performant way.
Posted
#research