Welcome to the Beam Summit 2023 Talks repository! This repository contains slides and resources from talks given during the bEam Summit in 2023.
-
Talk 1: Apache Beam and Ensemble Modeling: A Winning Combination for Machine Learning
- Link to Slides
- YouTube Link
- Summary: Discover streamlined machine learning workflows with Apache Beam and ensemble modeling in this talk. Learn how Apache Beam's flexible data processing and the RunInference API simplify complex tasks, offering efficient orchestration of machine learning steps within a single DAG. The README includes a practical example of an ensemble model pipeline for image captioning, featuring the BLIP and CLIP models.
-
Talk 2: Accelerating Machine Learning Predictions with NVIDIA TensorRT and Apache Beam
- Link to Slides
- YouTube Link
- Summary: A talk on integrating NVIDIA TensorRT with Apache Beam SDK to simplify data preprocessing and enhance machine learning model inference at scale. The integration aims to improve efficiency, reduce costs, and optimize GPU utilization. The talk includes an example using the RunInference API with TensorRT for a sentiment analysis model, demonstrating significant speed improvements. Benchmarks support the achieved performance enhancements.