This Fall, we are launching a new seminar course: Systems for Large Language Models (SysLLM) at the University of Saarland.
Course Details
- Format: Research seminar
- Students: 12 students
- Language: English
- Role: Co-instructor
Why Systems for LLMs?
Large Language Models have revolutionized AI, but deploying them at scale presents significant systems challenges:
- Memory constraints: LLMs require enormous GPU memory
- Inference latency: Real-time applications demand fast responses
- Training efficiency: Distributed training across many accelerators
- Serving at scale: Handling millions of concurrent requests
This seminar explores cutting-edge research addressing these challenges, from model parallelism and quantization to efficient attention mechanisms and serving systems.
Students read and present recent papers from top venues, developing critical analysis skills while learning about this rapidly evolving field.
For more information, visit the seminar website.