Legacy article

This post reflects an earlier stage of Empathy Platform development. Some of the tools, integrations, or approaches described here are no longer in use in our current stack.

Since then, our focus has evolved towards self-hosted, private, and sustainable AI infrastructure, where compute is treated as part of the product itself. All AI compute is currently run on Empathy’s own GPU environment, hosted in our net-zero energy, bioclimatic private cloud in Asturias.

Rather than updating this article to fit our current approach, we’ve chosen to preserve it as a record of our R&D history and innovation journey.

For our current approach and latest developments, explore our main blog section.

At Empathy.co, we continuously strive to push the boundaries of innovation, particularly in enhancing search experiences. Our recent project involved fine-tuning the Mistral model to improve content search experiences. Despite our focus applies to many content scenarios, these post series show how we improved our developer portal search functionality to provide a great example of what’s possible.

This post is the first in a four-part series, focusing specifically on the overall journey. The second post will provide a backend perspective, and subsequent posts will explore the infrastructure and user experience perspectives.

Fig. HolonSearch Privacy-First Generative Content Search experience in Empathy Platform Docs

The horizons of Generative AI

From the very beginning of AI irrupting in our daily lives, we can outline two horizons of generative AI experiences:

• Interactive query & retrieval experiences that involve engaging in a conversation where you ask questions and receive answers. The interaction is based on natural language input.
• Data-driven automation experiences that involve providing richer file datasets or data from emails, meeting notes, calendars, etc. So, AI-like tools can process this data to generate responses and automate tasks.

Following these approaches without caring about the importance of privacy awareness can lead to significant risks, including data misuse, erosion of user trust, and potential breaches of sensitive information. Our goal to elicit trust on our way to leveraging AI in search is to start with the aspects of an emotionally engaging experience, focusing on creating an intuitive and private user experience.

So, we’ve created a seamless AI content search experience that maintains people’s trust and privacy while interacting with it, encapsulated by our ethos of “We love AI. We love privacy.” By integrating a privacy-minded approach, we ensure a holistic and responsible development of AI technologies.

A unified vision

In developing this project, we aimed to create a seamless and intuitive search experience powered by generative AI while ensuring people’s trust and privacy. This integral vision required a coordinated effort across three main areas:

Backend development

Our backend teams focused on generating and fine-tuning datasets for the Mistral AI model. By creating diverse data sets and using advanced machine learning techniques, we ensured the model could provide accurate and contextually relevant responses.

Get to know more about the backend process in the second part of the post series.

Infrastructure

To support the enhanced AI capabilities, we designed a flexible and secure infrastructure. This involved using open standards to avoid vendor lock-in and implementing AI gateways to manage interactions between clients and multiple AI model providers. Our setup included both self-hosted and cloud-based solutions, ensuring scalability and reliability.

For more details on the infrastructure, check out the third part of this series.

UX and frontend development

The frontend team worked to translate the backend capabilities and the robust infrastructure into a user-friendly interface, upgrading people’s interaction and search experience. By integrating natural language processing into the existing contextual search function yet maintaining traditional keyword search options, we created an intuitive and familiar search experience, ensuring it met the needs of both beginners and experienced users.

Discover more on the user experience approach in the last post of the series.

The result: a unified search experience

The fine-tuning of Mistral for our developer portal exemplifies our commitment to innovation and user privacy. By combining diverse data sources, leveraging advanced AI models, and ensuring flexible deployment, we’ve created a powerful and private search experience.

Building search, differently

What started as early experimentation has evolved into a more integrated way of building AI search. Today, our work is centered around Empathy.AI, the space where we design and develop AI search systems grounded in self-hosted, private, and sustainable infrastructure.

By treating compute as part of the product itself, we gain greater control, efficiency, and long-term scalability.

Want to explore how this approach shapes what we build today? Discover more about Empathy.AI (opens new window) or dive into our latest articles.