Venu Tammali

About Venu Tammali

I like Mathematics, Computer Science and Coffee.

Experience

1. Sr Applied Scientist

   Amazon

   Oct 2021 — Present · Seattle, WA, US

   Graph Foundational ModelBuilt temporal graph foundational model (Temporal Graph Neural Network (TGN)) with graph size millions of nodes and billions of edges. Pre-trained the foundational model with link prediction task and built LGBM models that consume TGN model embeddings for several downstream tasks.Safe Reinforcement LearningProposed and built SafeRL to overcome the scenario of harmful actions suggested by RL agent which is trained offline and deployed online.Knowledge DistillationUtilized Knowledge Distillation on deep Temporal Self Attention (TSA) model to perform self-distillation to get a TSA model with same model size, but better prediction performance. Performed offline-distillation to distill a much smaller TSA model with similar prediction behavior. Besides distillation, also utilized dynamic quantization method to quantize the model, which significantly reduced the model size.

Education

• University of Illinois Chicago

  Doctor of Philosophy (PhD), Applied Mathematics

  2013 — 2015

• Osmania University

  Bachelor of Engineering (BE), Electrical, Electronics and Communications Engineering

• New York University

  Master of Science (MS), Applied Mathematics

  2010 — 2013