On October 2nd and 3rd, we hosted our 2024 edition of the Algolia DevCon, where Algolia’s vibrant community gathered to exchange ideas and learn from one another. This year’s theme was “Understanding the Spectrum of AI Search” and we did talk about all of that. We talked about understanding user intent, we talked about making AI understandable and we talked at length about search in all its specificity.

If you missed the event, don’t worry! Here’s a recap of the main themes and highlights, along with links to the recordings of every session. Whether you’re interested in technical deep dives, creative experiments, or personal stories from our team, you’ll find plenty to explore.

The only constant is change

The conference opened with Xavier Grand, sharing his ten year journey at Algolia, from intern to newly appointed CTO. Xavier reflected on the different phases of a startup's evolution, emphasizing the importance of adaptability and continuous learning, both at the company level and as an individual.

He spoke about his personal growth, transitioning from a focus on coding to managing and from hard skills to soft skills. As he put it, "There are many layers to working in a tech company. Coding is just one part of many. I was fortunate to grow alongside Algolia, and it taught me some invaluable lessons."

“If you can’t explain it, you can’t control it.”

Building on the theme of hard-earned lessons, Dustin Coates, with over a decade of experience working on search and AI systems at Algolia, emphasized the critical importance of explainability. He captured it succinctly: "If you can't explain it, you can't control it." His insight set the stage for a broader discussion about the responsibility to make AI understandable.

François Chalifour built on this idea by zooming in on personalization, a more specific application of AI. He warned of the risks of poorly implemented personalization, saying, "Bad personalization is worse than no personalization, as the joy of being listened to vanishes when we’re misunderstood." His talk underscored that transparency in AI isn’t just about functionality but also about respecting the user’s intent.

Maria Lungu’s presentation took a high-level step back to reflect on the broader principles guiding UI/UX in generative AI products. Drawing on a framework by IBM, she outlined foundational principles for creating trustworthy systems. Explainability is one of these principles, but it sits alongside others that are equally critical for supporting user trust and intent. Her talk walked us through those principles and how she and her team applied them in our upcoming auto-generated shopping guides.

User-centric search

In search technology, it's crucial to remember that the ultimate goal is to help the end user find what they are looking for. Yet, when deeply immersed in the technical details of building a search engine, it's easy to lose sight of this perspective.

That’s why I deeply appreciated the discussion between Charlie Hull, David Pugh, Daniel Tunkelang, and Etienne Martin. With decades of experience in designing and refining search systems, their insights were not only deeply interesting but also remarkably accessible, making this panel one of the highlights of the conference. I’m always impressed by experts that can turn complex concepts into simple explanations, while keeping the nuances intact.

One such example was their explanation of relevance versus ranking, which set the tone right from the start. "Relevance is like casting a fishing net into the water, aiming to catch as many fish as possible. Ranking, by contrast, determines which fish to bring to the surface first." I highly recommend watching the session's recording, as the panel was filled with insightful nuggets like this.

Easier data ingestion

DevCon was not only about high-level meta principles; we also went deeper into some specific Algolia features (some upcoming brand new beta features, and some older but maybe less-known ones).

Pushing data into Algolia is pretty straightforward. If you can get your data in JSON format, you’re just one API call away from having it in your index. But in practice, higher-quality search results come from higher-quality input data, so there’s often a necessary step of cleaning and transforming the data beforehand. Joni Khattar and Morgan Leroy introduced our new integration platform that simplifies this process by using your own JavaScript transform functions directly at indexing time, on Algolia’s servers!

These functions take raw data as an input and transform it into clean records right before it is indexed.. Whether it’s calculating discounts, assigning popularity scores, or enriching records with external API calls, these transformations happen seamlessly and in real time. We hope this small low-code enhancement will help simplify your ingestion pipelines.

The Algolia Crawler is another of our low-code platforms, and we’ve already introduced it in previous DevCon editions. This time, Tanya gave us a step-by-step guide of how to use it in practice. By extracting content directly from HTML pages and updating the index automatically, the Crawler provides an easy way to maintain accurate, up-to-date information without requiring access to the source code. As long as you have data accessible in HTML format somewhere online, the Crawler can get it and keep it in sync in your Algolia index.

The importance of user intent detection

While simplifying data integration makes life easier for developers, it’s important to remind ourselves once again that the ultimate goal is to create a search experience that truly satisfies the end user.

David Bacon's talk, delivered with humor and a flair for musical metaphors, highlighted an oldie but goodie feature of ours: Query Categorization. He began with an example many can relate to—searching for "4K TV" and receiving results for remote controls, adapters, and other accessories instead of televisions. While these results are technically correct, they may not align with the user’s true intent.

Query categorization solves this problem by predicting which categories should take priority based on search traffic patterns. As we know that when users search for the "4K TV" query, they usually are buying items in the “televisions” category, so we can prioritize results coming from that category. That’s what Query Categorization (or QCat for short), does. This feature includes configurable controls, allowing users to adjust how strongly the categorization influences results, either filtering on that category or simply weighting them more heavily. Once again, putting configuration in the user’s hand.

In a related session, Taylor Johnson demonstrated how Algolia’s Rules engine can extend search functionality through custom triggers and consequences. For instance, a specific query would trigger the display of a banner or custom HTML element directly within the search results. Allowing you in turn to change the user experience by inserting contextually relevant content, such as promotions or announcements, alongside search results.

Together, these features showcase Algolia’s ability to go beyond search, offering dynamic, context-aware solutions that adapt to user intent.

Algolia usage in the wild

That adaptability also inspired developers from outside of the company. For example, Matteo Rigoni, a developer from Milan, shared how his team needed both SSR (Server-Side Rendering) and Algolia’s search capability. While we still advocate by default for front-end search, we know that there are legitimate use cases for SSR (most notably better SEO on product pages).

Matteo demonstrated how his team built a custom Nuxt plugin, SwiftSearch, to have Algolia InstantSearch work with SSR. It was fascinating to observe how our libraries, despite not handling a particular use case by default, could still be tweaked and adapted by using the various APIs and hooks we offer. Well done, Matteo!

Next, Raymond Camden, a seasoned blogger and longtime Algolia user, provided a more personal perspective by showing how he integrated Algolia's Recommend feature into his blog. While designed primarily for e-commerce, Algolia Recommend was adapted by Raymond to suggest related blog posts, allowing readers of his blog to find more interesting posts from his prolific writing.

With more than 6000 (!) posts on his personal blog, cost optimization was crucial for Raymond; he didn’t want to spend too much money on a recommendation system. So he showed us the tricks and hacks he used during his implementation. I loved that presentation for the spirit of experimentation he showed, demonstrating how small adjustments and creative thinking can make a significant impact.

Lucas Bonomi, our resident CSS Crazy Expert, delivered an even more unconventional and wild presentation. He found a way to store and play videos using CSS only, using and abusing box-shadow syntax and animations. And if that wasn’t crazy enough, he even managed to store those CSS bits in Algolia, turning his index into a database for storing and playing videos.

You might wonder what the practical use of this approach could be—and that’s exactly the point. Sometimes, it’s not about the end result but the journey of experimentation and discovery. By pushing the boundaries of what a tool can do, we learn, innovate, and open the door to new possibilities we might never have imagined.

The Algolians: the beating heart of Algolia

Building outstanding products goes far beyond technical expertise—it’s about the people behind the code and the challenges they face every day.

Ben Greenbank illustrated this perfectly by sharing his journey of overcoming imposter syndrome and his struggles with remote work. His talk stood out to me, a reminder that we can learn at any stage of our careers, from colleagues both junior and senior. Ben’s candid reflections on seeking help, connecting with others, and acknowledging cultural differences offer lessons that resonate far beyond the world of coding. If you haven't already, I highly recommend watching his talk: Amplify your Impact, without writing any code.

If Ben’s story focused on self-reflection and personal growth, Stefan took us into the study of groups and how they interact. He shared how his background in anthropology shapes his role as a technical program manager (TPM) and how he sees himself as an Engineer Whisperer. Stefan is part anthropologist, part mediator, working to bridge the gaps between engineering, product, and sales teams. By understanding each group’s culture (their jargon, rituals, and beliefs), Stefan fosters collaboration while protecting engineers from their kryptonite: constant interruptions, shifting scopes, and impossible deadlines.

In keeping with the October theme of this DevCon, Emmanuel then shared a chilling, Halloween-themed "tech horror story" about a significant production incident that occurred early in his career. Bugs, debugging chaos, and an e-commerce mishap straight out of a nightmare—all of that in production. It was the kind of scenario every developer dreads, yet one that even the most experienced among us have faced at least once.

Those stories are just a glimpse of the diverse and fascinating backgrounds that Algolians bring to the table. This Algolia DevCon 2024 was packed with incredible talks, and while this blog post can’t cover them all, I’d like to quickly highlight a few more experts who shared insights on topics close to their hearts.

From J Choi, discussing the dual role of the good bots (like our Crawler) and the bad bots (the spammy kind), and how to protect your Algolia index from the latter. Tom Clarke provided practical advice on FinOps, advocating for a clear understanding of cloud costs and shared vocabulary (namely the ubiquitous “unit cost”). Finally, Paul-Louis Nech hosted a panel about Retrieval Augmented Generation (RAG) and explored the challenges of both protecting sensitive information while leveraging AI innovations.

So Long, and Thanks for All the Searches

We hope this recap gave you a glimpse into the richness of DevCon and inspired new ideas for your projects. DevCon is our annual celebration of innovation, and we’re already looking forward to next year’s edition. In the meantime, we’ll be hosting smaller quarterly DevBit events soon—stay tuned! If you’ve implemented something exciting with Algolia or have topics you’d love us to explore in future events, let us know. We’d be thrilled to highlight your work or dive deeper into areas that matter most to you.