Quick to Production: Integrating Spark and TensorFlow for Efficient MLOps

Discover how to rapidly deploy deep learning and machine learning models using TensorFlow in a production environment immediately after prototyping. Learn to leverage the strengths of both Spark and TensorFlow in a single project, including TensorFlow ecosystem libraries like TensorFlow Hub, TensorFlow Recommenders, and ranking. Explore techniques for handling big datasets in a distributed setting with minimal MLOps code, allowing data scientists to focus on feature engineering and model building. Gain insights into simplifying the process with Databricks handling most of the MLOps, enabling small teams to efficiently work with TensorFlow in distributed environments. Understand the benefits of combining Spark and TensorFlow solutions, including batch inference, experiment tracking, model management, and serving endpoints. Delve into topics such as feature engineering, data distribution, TensorFlow Extended, and the advantages of using Spark libraries and pipelines.

Introduction
Welcome
What is this topic
What we will cover
Who is this talk for
Why do we need deep learning
Deep learning solutions
Who we are
Previously
Feature Engineering on TensorFlow
Benefits of Spark
Spark Libraries
pandas udif
Spark ML
Spark Pipeline
Questions
TensorFlow
TensorFlow Data
TensorFlow Distribution
Data Distribution
TensorFlow Extended
TensorFlow Transform
TensorFlow Recommenders
TensorFlow Hub
Batch vs RealTime
Experiment Tracking
Model Management
Serve Models
Serve Endpoint
Combine Solutions
TensorFlow Record
Spark Library
TensorFlow Distributor
TensorFlow Distributor Code
TensorFlow Distributor Nodes
Saving Models
Batch Inference
Recap
Pros
Challenges
Data and AI Summit