Platform-First AI: How We Built a Secure AI Infrastructure and Showcased It Through Drupal
Time: Thursday 14th May 13:20-13:50
Room: Loft
Track: Sponsor talk
Most organizations approach AI integration by starting with use cases and then scrambling to find infrastructure that meets their security and compliance needs. We took the opposite approach: we built a comprehensive AI platform first, then developed a suite of Drupal-based tools and modules to demonstrate its capabilities.
This talk will explore how amazee.ai created (and open-sourced) an AI infrastructure, designed around LiteLLM, to support multiple AI providers while maintaining complete data sovereignty. We'll discuss the key architectural decisions that went into building a platform that handles authentication, vector databases, and multi-region deployment - all while ensuring no data is stored or used for model training.
Then, we'll showcase how we leveraged this platform to develop practical Drupal integrations, including:
* A custom Drupal AI Provider module that enables one-click setup without managing API keys
* Default integration with Drupal CMS 2.0 and the Drupal AI module ecosystem for content generation, accessibility improvements, and multilingual support
* Real-world implementations solving challenges like batch alt-text generation for 10k+ product catalogs, interactive chatbots powered by vector databases and more!
* And we’ll even cover how (and why!) we used Drupal as the base for our private AI Assistant
Attendees will learn why choosing and building infrastructure first - rather than bolting it on later - supports more flexible, secure, and scalable AI solutions. We'll share lessons learned from our platform-first approach, including technical decisions around user management & reporting, regional compliance, and provider-agnostic architecture.
This talk will explore how amazee.ai created (and open-sourced) an AI infrastructure, designed around LiteLLM, to support multiple AI providers while maintaining complete data sovereignty. We'll discuss the key architectural decisions that went into building a platform that handles authentication, vector databases, and multi-region deployment - all while ensuring no data is stored or used for model training.
Then, we'll showcase how we leveraged this platform to develop practical Drupal integrations, including:
* A custom Drupal AI Provider module that enables one-click setup without managing API keys
* Default integration with Drupal CMS 2.0 and the Drupal AI module ecosystem for content generation, accessibility improvements, and multilingual support
* Real-world implementations solving challenges like batch alt-text generation for 10k+ product catalogs, interactive chatbots powered by vector databases and more!
* And we’ll even cover how (and why!) we used Drupal as the base for our private AI Assistant
Attendees will learn why choosing and building infrastructure first - rather than bolting it on later - supports more flexible, secure, and scalable AI solutions. We'll share lessons learned from our platform-first approach, including technical decisions around user management & reporting, regional compliance, and provider-agnostic architecture.
Speakers
Toby Bellwood
Toby is the Head of Engineering at amazee.io, where he leads the engineering organization and oversees development of Lagoon, the company's open-source application delivery platform, as well as the Platform Engineering and AI engineering teams. Since joining in early 2020, Toby has shaped the technical direction of amazee.io's products, focusing on developer experience, platform architecture, and continuous security improvements. His extensive experience in open-source, cloud, and open-data communities brings a unique perspective to building platforms that prioritise security, sovereignty, and accessibility.
Before joining amazee.io, Toby spent many years in the Australian public service as technical lead for GovCMS, data.gov.au, and NationalMap. This experience gave him deep insights into the importance of open-source solutions, data sovereignty, and building platforms that serve diverse user needs.
Toby has spoken at numerous events across the open-source, open-data, and open-government spaces—including DrupalSouth, DevOpsDays, Everything Open, and various Kubernetes conferences. His presentations cover topics from application-specific implementations to cloud-centric architecture, developer experience, and organizational culture. He's passionate about making complex technical topics accessible and building products in the open with community participation.
Before joining amazee.io, Toby spent many years in the Australian public service as technical lead for GovCMS, data.gov.au, and NationalMap. This experience gave him deep insights into the importance of open-source solutions, data sovereignty, and building platforms that serve diverse user needs.
Toby has spoken at numerous events across the open-source, open-data, and open-government spaces—including DrupalSouth, DevOpsDays, Everything Open, and various Kubernetes conferences. His presentations cover topics from application-specific implementations to cloud-centric architecture, developer experience, and organizational culture. He's passionate about making complex technical topics accessible and building products in the open with community participation.