Spring is here, and with it comes another installment of Algolia’s quarterly livestream event, DevBit. This quarter, we focused on recommendations — how to use data from your index records and user analytics to craft media and product suggestions. We call this “putting your data in motion.”

If you missed the event (or just want a refresher on what was presented), this post is for you. I’ve included short recaps and links to the recordings of all the sessions. There’s also a playlist on YouTube. We had three sessions this quarter, including our very first-ever panel discussion!

• Building for Discovery: A journey into Recommendations
  by Lucia Kubinska, Senior Product Designer | Recommend at Algolia
• Events: what to capture and why?
  panel discussion with Samy Jarboui, Ruxandra Voicu, and Chuck Meyer | Algolia
• Visual recommendations: the Good, the Bad, and the Ugly
  by Paul-Louis Nech, Senior ML Engineer | GenAI & Recommend at Algolia

Building for Discovery: A journey into Recommendations

Lucia Kubinska brought her unique perspective as a senior product designer to the topic of recommendations. Her presentation described the different recommendation models and how to best use (or avoid) them when designing media and e-commerce user experiences. She also helped us understand the nuances between recommendations and personalization

Events: what to capture and why?

Next was our first ever DevBit panel discussion. I chatted with senior machine learning engineer Ruxandra Voicu and solution architect Samy Jarboui about using click and conversion event data to train model and make more informed recommendations for your users. I learned that a lot has changed in the last year and there are now several recommendation models you can use with minimal, or even no user data. My biggest take away — sometimes bad data is worse than no data at all. Thanks to all of the attendees who contributed live questions and commentary!

Visual recommendations: the Good, the Bad, and the Ugly

A picture is worth a thousand words, so what can you do with a thousand pictures? Paul-Louis has spent most of the last year training our “Looking Similar” models for image recommendations. He brought that deep knowledge to the stage with a presentation on various strategies around training models using images. He showed some approaches that bordered on magical, and some that were pretty much the opposite.

After the event we all adjourned to the Algolia Developer Discord server to answer questions and chat about the sessions. You can come join us there too!

And that was Spring 2024. Thank you to everyone who attended live and everyone watching now. We’ll see you in June for the next Algolia DevBit!