Kousik Rajesh
Employers on the program
- 2025Pinterest1 talk
Events1
Conferences1
Talks1
Years active2025
Appearances by year
- PyTorch Conference
Talks
| Date | Event | Conference | Role | Title | Links |
|---|---|---|---|---|---|
| Oct 22, 2025 | PyTorch Conference 2025 | PyTorch Conference | speaker | Scaling Inference of O(10K)-length Sequence Recommendation Models Using CudaGraph and Triton Kernels as Pinterest In this talk we discuss how we scaled Pinterest’s RecSys models from a context length of 500 to 16000 for a 10M+ QPS system while maintaining latency and…Show lessIn this talk we discuss how we scaled Pinterest’s RecSys models from a context length of 500 to 16000 for a 10M+ QPS system while maintaining latency and throughput. We start with how we optimized goodput, using Torchscript on a custom CudaGraph backend and OpenAI triton kernels to fuse expensive... |
Frequent co-speakers
| Speaker | Shared events | Shared talks | Last together |
|---|---|---|---|
| Saurabh Vishwas Joshi | 1 | 1 | 2025 |
Speakers with the same name are merged into one entry, so this page may combine more than one person. Affiliations are shown as each conference listed them.