This is Yubo Wang, a software engineer focused on AI infrastructure. I currently work at Together AI on inference engines. Previously, I led the development of LinkedIn’s AI Training Platform, where my team leveraged Flyte to support diverse training workloads including RecSys models and LLMs.

My latest public work is TorchSpec, a torch-native framework for disaggregated speculative decoding training. We wrote about it in the PyTorch blog post TorchSpec: Speculative Decoding Training at Scale.

I have a 8-year-old cat named 鼻妞. She kneads really good dough.