Ketan Singh, software engineer at Meta, will deliver "Trends in Scaling GenAI Workloads" as part of the IST Colloquium Talks series. The event will be held virtually via Zoom and is open to the public.
About the Talk
Generative AI (GenAI) models are transforming industries, but their massive size and computational demands pose unique scaling challenges. This talk will explore the latest trends and techniques for effectively scaling GenAI workloads to unlock their full potential, including:
- Hardware Innovations: the role of specialized accelerators (GPUs, TPUs, etc.) and how they're driving performance gains
- Distributed Training: strategies for parallelizing training across multiple devices and data centers to handle massive datasets
- Optimization Techniques: methods, from quantization to pruning, to reduce model size and boost efficiency without sacrificing accuracy
The talk will provide insights into the cutting-edge approaches for scaling GenAI (and other applications) and maximizing its impact.
About the Speaker
Ketan Singh is an accomplished technology leader with a proven track record of innovation in machine learning and artificial intelligence. After graduating from the Indian Institute of Technology Guwahati and earning his master's degree from USC, he honed his skills at industry giants Apple, Google, Stripe, and now Meta. Singh has pioneered patented technologies in navigation, proactive features, search ranking, and fraud detection. Currently, he leads Meta's Ads Model Serving initiatives, driving large-scale ML infrastructure and shaping the future of sequence modeling in the industry.