Bio
Yunmo Koo is a founding engineer at FriendliAI focused on LLM inference optimization, distributed training, multi-cloud systems, and LLMOps. He builds production ML infrastructure for lower latency, better reliability, and improved cost efficiency.
Session (1)
Additional Speaker Details
Using publicly available information we constructed an analysis to help you get a feel for this speaker before deciding to attend their session.
Yunmo Koo is a Founding Engineer at FriendliAI who has spent over five years building the core LLM inference runtime — speculative decoding, KV-cache optimization, GPU kernels, and continuous batching — and now applies that deep systems knowledge to the demands of agentic AI. His session offers first-hand insight into what it takes to make multi-step agent workloads fast and cost-efficient in production.
Speaking style
Recent talks (2)
GitHub
@kooyunmo- kooyunmo/GAN-detector★ 22Fake image detection — identifies which GAN generated a synthetic image
- kooyunmo/cuda-uvm-gpt2★ 17PyTorch implementation using CUDA unified virtual memory for large language models
- kooyunmo/pytorch-uvm★ 15Fork of PyTorch with GPU unified virtual memory acceleration modifications