Scale from thousands to millions of requests and AI generations monthly with Firebase
Maintain low latency and downtime with Cloud Run
Handle over 100 TB of generated images and assets with Cloud Storage
Generate insights and data analysis of over 25M users using BigQuery
Optimize management infrastructure with a small team of four
With the help of Google Cloud solutions, NightCafe is helping people enjoy being creative with an extensive range of AI models, along with the advantages of a robust IT infrastructure and growing community.
With the help of Google Cloud solutions, NightCafe is helping people enjoy being creative with an extensive range of AI models, along with the advantages of a robust IT infrastructure and growing community.
Empowering creators and growing a community with AI
Pretty much all the services we use on Google Cloud are fully managed, so we've never really had infrastructure scaling issues. However, over time, there have been some things we need to refactor, or do a different way entirely at scale, and those have always been manageable.
Angus Russell
Founder, NightCafe Studio
As a process, art creation typically requires skill, time, and effort, leaving many unable to experience the satisfaction of seeing an idea come to life. Since 2019, NightCafe Studio has been a platform that enables users to express their creativity with AI, regardless of their skill level. Moreover, with NightCafe being a low-capital startup competing against much bigger, established players, Google Cloud's Managed Services helps the company level the playing field, and innovate and scale without significant outside investment.
With a focus on the community, the platform provides an extensive range of AI models, community features, gamification, AI art challenges, and AI-generated prompt suggestions to help people enjoy this hobby and overcome creative blocks.
The onset of the generative AI boom over the past few years meant that NightCafe would soon welcome a surge in users and requests. Built on a foundation of Google Cloud and Firebase platforms, developers can rely on a range of infrastructure and Artificial Intelligence (AI) capabilities. This meant that even a small engineering team of just four people could continue to handle growing demands.
"Pretty much all the services we use on Google Cloud are fully managed, so we've never really had infrastructure scaling issues. However, over time, there have been some things we need to refactor, or do a different way entirely at scale, and those have always been manageable," shared Angus Russell, founder of NightCafe Studio.
"We have tens of billions of records in Firestore, tens of millions of authorized users, more than 100 TB of cloud storage, and over 100M cloud function invocations per day, and none of them have ever really missed a beat."
To date, the platform has scaled from handling thousands to millions of requests and AI generations monthly while maintaining low latency and downtime – this reflects growth enabled by the platform rather than through migration improvements.
Lean stack for massive scale
For Russell and his team, it is vital to keep the technology stack as simple as possible, as they want to remain a bootstrapped entity competing against well-funded competitors. This means having as many managed components as possible with dedicated features that can enable growth.
The company's initial use of Firebase was driven by its seamless integration and speed to market, and the team has continued to expand the use of other Google Cloud solutions throughout the tech stack to address additional needs as business grows.
"As we've scaled to over 1B images generated by 25M users, our initial Firebase-centric stack has continued to meet our needs. While we've addressed inefficiencies from our early side-project phase, Firebase's capabilities and advanced usage patterns have grown with us, allowing us to scale without requiring a transition to a more traditional enterprise infrastructure," said Russell.
We expected to need dedicated infrastructure or DevOps resources as we scaled, but the managed nature of services like Cloud Run functions and Firestore meant we could focus our engineering efforts on building AI features rather than managing infrastructure.
Angus Russell
Founder, NightCafe Studio
Using Firestore as its primary database, NightCafe Studio has been able to scale smoothly, handling billions of read and write operations daily while keeping the reactive user experience that is the standard. Over hundreds of terabytes of user-generated images and assets are being stored securely on Cloud Storage, with Memorystore for Redis serving as a caching layer to maintain low-latency data access with high throughput across the entire experience, both internally and externally, and helping to reduce costs as queries increase.
Workflows behind the scenes have been enhanced with the use of Cloud Run functions, which provide a fully managed, auto-scaling setup that increases reliability for both event-driven workflows and direct API requests. This allows the team to focus on building features rather than managing infrastructure.
"We expected to need dedicated infrastructure or DevOps resources as we scaled, but the managed nature of services like Cloud Run functions and Firestore meant we could focus our engineering efforts on building AI features rather than managing infrastructure," added Russell.
Enabling resilient service and data monitoring
As a consumer-facing service, it is imperative that the platform remains functional whenever the urge to get creative arises. The NightCafe team relies on Cloud Monitoring to oversee the entire tech stack, with alerts and logs providing full observability in the event of any unforeseen issues.
"As we matured, we also started using monitoring and alerts. Cloud functions actually have pretty good built-in monitoring dashboards for things such as errors, memory usage, instance count, etc. But we now also use monitoring for dashboards that show usage, error rates, latency, and more for each model we run on NightCafe. We can also break the dashboards down by model, API provider, etc.," shared Russell.
At the same time, NightCafe's developers have also set up alerts for error rates on the models used, with some of the higher-volume models having a fallback API provider if the primary provider goes down. If the alert triggers for those models, a feature flag is automatically activated to divert traffic to the backup provider, ensuring uninterrupted access to create.
The large amount of data available is also effectively utilized with BigQuery and Pub/Sub. The latter serves as a secure, scalable message queuing system with access controls and always-on encryption for the former, providing developers the opportunity to analyze data, build reports and insights securely and accurately.
Tapping Vertex AI, using AI for good, and fostering a positive community
We leverage Vertex AI for image generation using models such as Imagen 3, and LLM-powered text generation for critical workflows such as content personalization and automated summarizations. This AI integration has significantly improved our product capabilities while streamlining operations.
Angus Russell
Founder, NightCafe Studio
The advantages and benefits of generative AI are evident not just for the users but also for the team behind NightCafe Studio.
"We leverage Vertex AI for image generation using models like Imagen 4, and LLM-powered text generation for critical workflows like content personalization and automated summarizations. This AI integration has significantly improved our product capabilities while streamlining operations," added Russell.
The seamless and straightforward integration into the existing tech stack using Vertex AI provides the team the opportunity to focus on curating the best AI experience for the users rather than managing complex integrations. Over time, new models and providers have also been added to the offerings, giving users an even greater degree of choice.
However, the creation of AI art has always sparked discussion, and it is a conversation in which NightCafe Studio is eager to participate.
"On NightCafe, our users are just here to enjoy the process of being creative. They're not replacing traditional artists in any sense," Russell said. "It's absolutely fair to be worried about people's jobs. But I also see the feedback from NightCafe users who feel happier after being creative, who use AI Art for therapy, who have found a hobby that they love, and who are not displacing anyone by creating art for fun."
The team will continue to help users discover the joy of creating, and to build features which actively enable that. At the same time, it will utilize layers of protection such as AI-powered prompt moderation, AI-powered image moderation, and more to moderate and protect, creating – and building – a more vibrant, positive community in the industry.
NightCafe Studio is a leading AI art generation platform that empowers millions of users to explore their creativity. As a bootstrapped company, they have successfully competed with heavily funded competitors by focusing on community, innovation, and a lean, scalable infrastructure powered by Google Cloud.
Industry: Media & Entertainment
Location: Australia
Products: Google Cloud, BigQuery, Cloud Monitoring, Cloud Run, Cloud Storage, Firebase, Identity and Access Management (IAM), Memorystore for Redis, Pub/Sub, Vertex AI
