We host ou starter to pro plans on high-end clusters, we usually host up to 100 customers on a cluster. When we install a new cluster, it is empty. Of course users can grow above the limit of the cluster, in this case we migrate some users on the new cluster, this is an operation which is transparent for our users and let them upgrade or downgrade. We ensure we always have more resource than the current usage (we keep around 25% of free resource on our global infra for scaling) For our big enterprise customer, we have a dedicated infrastructure for them and they have the control on upgrade/downgrade. We also have a migration procedure to move from a pro plan to an dedicated infrastructure, you don't need to reimport your data. With the growth we have, we add new machines in production close to every day. So we scale up the infra very often. There is no technical limitation to scale down (same principle than scaling up), we hope it will never be needed :)

Elastic is a trendy keyword that can means a lot of things. We are elastic in several ways: we can add more indexing capacity (more clusters) we can add more search capacity (more machines with copy of data, our biggest user perform around 150M queries per day) we can distribute your search worldwide while ensuring a fast search everywhere (adding DSN replicates in a different region) we have all data in memory in 99% of cases but we have few use users where over-booking make sense to reduce costs (over-booking of memory) At the end, elasticity depends a lot of the use case

Culture fit first, we like people that want to think about the most important thing for the company (we are 100% transparent on all topics). Then expertise, as a team we search for people that are better than us on at least one topic and that bring something new. The hiring process includes one day in the team with a final presentation on a topic you are passionate about.

Yes building a game based on a search engine is doable. We have done one experiment with Imagga image tagging few months ago: https://blog.algolia.com/robots-vs-humans-who-will-emerge-supreme-in-the-battle-of-the-image-tags/

We also always have a huge travel budget plus a corporate flat in Paris and SF. Everyone in the team go to the remote office on a regular basis

There is no budget for performance, it is a daily focus. We know how easy it is to introduce a regression of performance and how hard it is to always improve it, we work on profiling and benchmarking nearly every single day!

45 and we plan to double the size of the team before end of the year :)

Our first offline SDK is still used in several mobile apps. I remember one benchmark with one index of 1M record having search performance under 50ms on a old Iphone 3GS. Seems so old right now ... but we have kept most of the algorithms and datastructures in our SaaS API.

We have more than 3000 active applications using us. We process more than 30 billion API calls per month. Algolia customers include: Vevo, Periscope, Quiksilver, Medium, Product Hunt, A Little Market (an Etsy company), CrunchBase, HackerNews, Genius.com and Livestream. We unfortunately cannot comment on our biggest customers because of NDAs... this is so sad :(

Our website is hosted on AWS, and we rely on a highly available MySQL (using RDS). Our API uses our own binary data structure. We have built everything with efficiency in mind, maximizing usage of CPU-cache, in order to reply in a few milliseconds.