Using WebAssembly for in-database Machine Learning

Eliminate data movement bottlenecks by running ML models inside your database. Learn how WebAssembly provides a secure, high-performance solution.

#1about 4 minutes

Introducing WebAssembly for in-database machine learning

WebAssembly enables running high-performance code like C++ or Rust directly inside a database, co-locating analytics with data.

#2about 3 minutes

Comparing methods for machine learning with databases

An overview of different approaches to database ML includes using Apache Spark, Python libraries, built-in vector functions, and OpenAI embeddings.

#3about 7 minutes

Why use WebAssembly for in-database analytics

Running Wasm UDFs inside the database avoids data movement, extends DBMS features, and executes sandboxed code at near-native speed.

#4about 4 minutes

Setting up the Wasm and Rust development environment

A step-by-step guide to installing the necessary one-time dependencies, including the Wasmtime SDK, Rust toolchain, and wasm-bindgen.

#5about 4 minutes

Building a Wasm sentiment analysis function in Rust

The process involves creating an interface definition file (.wit), managing dependencies with cargo, and writing Rust code to wrap a sentiment analysis library.

#6about 5 minutes

Compiling and deploying the Wasm UDF to the database

After compiling the Rust code into a Wasm module, a specific tool is used to upload and register the function within the database system.

#7about 6 minutes

Live demo of sentiment analysis on a movie dataset

A live demonstration shows how to execute the deployed Wasm UDF with a SQL query against a table of IMDB movie reviews.

#8about 26 minutes

Summary, resources, and audience Q&A

A recap of Wasm's benefits for database extensibility is followed by links to resources and an audience Q&A session on implementation and performance.