The Flashcard’it Project: oXya’s Role in the Google AI Hackathon

Last April, the oXya Canada team joined the Google AI Hackathon, which this year focused on generative AI, specifically Google’s new AI model, Gemini. This 2024 edition challenged participants to create innovative applications using the capabilities of this advanced model.

Project Inspiration

After deciding to participate in the AI Hackathon, our team explored ideas to make full use of Gemini’s capabilities. The vision was clear: design an application to generate flashcards, offering a user-friendly tool to enhance learning and certification preparation within the workplace using generative AI.

The Flashcard’it project was born—an application designed to make learning engaging and speed up test and training content creation.

Leveraging Gemini in Flashcard’it 

With Gemini integrated, Flashcard’it leverages two flashcard generation methods. The first utilizes Gemini’s base knowledge, independent of any corporate data. The second method allows users to upload information in PDF form, enabling Gemini to generate more specific, contextual responses. Of course, there’s more complexity, but we’ll discuss that shortly.

In practice, users can generate flashcards by setting prompts, such as:

“Create a set of 20 questions with 4 possible answers each, marking the correct one with a star”

This setup enables flashcards for training, certification, or general learning on various topics, all presented in a quiz-like format.

Adding Context with RAG 

To create more tailored questions and answers, the AI model can be supplied with additional context, like internal training materials or company-specific documents. This proves especially valuable when designing training content for unique business topics.

Fine Tuning is one option, allowing a generative model to specialize by adjusting its parameters with focused data, enhancing accuracy while needing significant resources and careful data management to avoid biases.

Another approach is Retrieval Augmented Generation (RAG), which merges information retrieval with text generation. This offers adaptable responses without extensive retraining, though it relies on the quality and relevance of accessible information.

How does it work? First, company documents are organized into a corpus, divided into chunks, converted into vectors, and stored in a vector database. When a Flashcard’it query is submitted, the RAG model turns it into vectors, locates relevant documents, and the large language model (Gemini) integrates the query with retrieved content to create a final answer.

This method helps minimize the risk of generative AI “hallucinations” by anchoring responses in verified data, lowering the risk of generating incorrect or invented information.

The Future of Flashcard’it at oXya

The hackathon gave us an ideal setting to brainstorm and explore fresh ideas for generative AI integration. Flashcard’it’s future at oXya looks promising, with plans to launch the application internally as early as next autumn.

Following the hackathon, the team has been refining the prototype to suit corporate environments and the specific needs of our employees.

Of course, concerns on data security remains a top priority. While Flashcard’it currently targets non-sensitive data, we’ve given some careful consideration to these concerns, the hackathon provided a valuable opportunity to assess various security considerations.

Expanding AI Skills: oXya’s Broader Vision

The Google AI Hackathon aligns with oXya’s wider strategy to explore advanced technologies as part of our managed services. We’ve previously incorporated generative AI for tasks like calculating satisfaction scores, and developing this new application marks an important step forward in expanding our expertise to better serve and advise our customers.

Beyond just technical progress, last April’s event was a valuable opportunity to collaborate, exchange ideas, and build stronger team connections. Every team member contributed their perspective to push the project forward, fostering a true spirit of teamwork. You can learn more about the project in the video presentation prepared for the hackathon.

By participating in events like the Google AI Hackathon, we’re positioning oXya teams at the forefront of the latest AI advancements, all while closely monitoring security and confidentiality. Want to discover more about our innovative methods? Reach out to us here!