Our AI journey
Stack Overflow’s journey to AI
The journey to incorporating AI into our internal tech stack and product offerings began with a desire to make learning easier and more efficient for our global user base. Every day, Stack Overflow helps engineers, developers, and technologists of every kind learn by answering their questions or guiding them to discover new solutions and fresh approaches.
We know that learning is integral to this work—and that it’s hard. Especially for beginners, it’s hard to know where to start, which questions to ask first. After all, you don’t know what you don’t know, especially when you’re new to a topic. We also know that our Stack Overflow for Teams customers are looking for innovative ways for their teams to find information faster and collaborate more seamlessly.
That’s why we’ve set out to build a better search experience and an online course recommendation system, both powered by AI, and are adding exciting new AI capabilities to Stack Overflow for Teams with OverflowAI.
A better search experience
You can’t talk about learning without talking about search. It’s been central to the Stack Overflow experience from the beginning, when it was one of the first user features we built. We knew leveraging the power of AI would make the search experience more powerful, more personalized, and more illuminating than ever. That meant implementing semantic search on Stack Overflow, a project you can learn about in more depth here.
Semantic search, as opposed to lexical or keyword-based search, converts content into numerical vectors based on meaning or sentiment assigned by machine learning, allowing users to search using natural language instead of through rigid keywords.
An intuitive recommendation engine
AI isn’t only transforming the search experience on Stack Overflow; it’s also making it faster and easier for users to find structured learning opportunities that can advance their skill sets.
We built a cloud-first, content-based recommendation system to suggest online courses selected based on the content a user is currently perusing and provide them with a structured learning path relevant to what they want to know.
Even as we were developing this online course recommendation system, we were launching a new, modern data platform for Stack Overflow that would give us a fresh infrastructure, limitless options in terms of our tech stack, and a chance to start using state-of-the-art machine learning techniques.
Read more from our senior data scientist about how we built the course recommendation engine.