Uber Eats Personalized Recommendation Machine Learning

End-to-end led to building Meta first terabyte(TB) scale large model for personalized recommendation (including model building, model training & fine-tuning based on model parallel and data parallel, model compression, model inference, real-time continual fine-tuning, model evaluation and large model related tools, such as distributed feature importance and model understanding).Worked on Meta Pytorch GPU based large model distributed offline training, fine-tuning, model evaluation, real-time continual fine-tuning, model compression, distributed inference.Worked on Meta large scale caffe2 distributed training and distributed inference, etc.

Worked on the file system and network.

Worked on the file system and network.