SPEAKER 2026
Sakshi Tyagi
Senior Apllied Scientist, Thomson Reuters
About Talk
Scaling LLM Training Beyond Data Parallelism: The Open-Source Stack from PyTorch to Megatron
Choose the right open-source scaling lever (DDP → ZeRO/FSDP → tensor/pipeline parallelism) based on their actual bottleneck : memory, communication, or model size, instead of adding GPUs blindly.
TRACK: AI & ML: Opening Up the Tech
7th Oct 2026 | Hall C | Time:04:45-05:15
About Speaker
Sakshi Tyagi is a Senior Applied Scientist at Thomson Reuters, specialising in Generative AI, Agentic AI, multimodal AI, and accelerated computing. She focuses on turning AI research into scalable, production-ready solutions across healthcare, life sciences, robotics, and enterprise applications. Sakshi is also a passionate technology evangelist who regularly shares her expertise through keynotes, workshops, panels, and mentoring.