The goal of this week was for KatholiekOnderwijs Vlaanderen (KatOVla) to develop a AI-driven chatbot.
It was our job to use that week to:
Overall, the workshop aimed to set a solid foundation for the project, ensuring that everyone involved had a clear understanding of the goals, tasks, and technologies to be used.
My first task was to dive into Azure AI and REST APIs, which was crucial for understanding how we would set up the architecture to meet Katholiek Onderwijs Vlaanderen's needs. I got a bunch of links and videos about Azure AI Foundry and how we would set up the architecture. It was fascinating to learn how this technology would play a role in the project and how we would integrate it to achieve our goals. This exploration and testing of Azure AI directly supports the goal of exploring and testing various technologies to ensure they meet the project's needs.
The next morning, I was able to test out the Copilot Studio environment. I got to prepare a demo for the IT team at Katholiek Onderwijs Vlaanderen. I spent the afternoon testing Copilot and even created a chatbot within Teams for RoboRana. It was a cool experience, especially seeing how quickly I could set up something functional with Copilot. In the evening, I looked for more ways Copilot Studio could add value for Katholiek Onderwijs Vlaanderen.
Wednesday was all about the demo for the functional analysts, Greet and Magda. I put the finishing touches on the chatbot and expanded its functionalities with Generative AI. The demo went well, and they said it was an excellent presentation, which made me happy. The testers came by to try out the developers' chatbot, while I compared the chatbot's questions and answers with the desired answers from the analysis team. It was the first time using the Ragas library for this task. Ragas is a tool I use to evaluate the quality of the responses from the chatbot when asking specific questions. It helps determine whether the chatbot is functioning correctly and providing accurate answers based on the given questions. [DL2] It was challenging, but I quickly understood how it worked and processed about 100 questions and answers to calculate relevance and similarities.
Thursday's morning meeting went smoothly. Everyone knew what they had to do, and I continued working with Ragas, but I hit some obstacles. After some trial and error and help from my colleagues, I knew we had to collect more data. So, I got a new task: testing the chatbot by entering questions and transferring the answers to Excel. It felt good to complete this task, even if it meant helping my colleagues with their challenges. It was a team effort, and I realized that perseverance made a difference.
Friday was mainly a day for wrapping things up. I continued working on the list of questions and answers and prepared for an important demo for the executives. After the demo, we had a reflection meeting where we discussed progress and areas for improvement. That feedback was valuable and helped us understand where we should head with the project. At the end of the week, I documented everything in a weekly report, and we said goodbye to the colleagues we had gotten to know well over the past few days.
My first week at RoboRana exceeded my expectations. The support from my colleagues and mentor was fantastic, and I felt like part of the team. It was also special to contribute to a client-organized hackathon, where we could create something that added value. All of this made me look forward to the coming weeks and the new challenges we'll tackle. I want to thank everyone who helped me: the IT team at Katholiek Onderwijs Vlaanderen (Reg, Magda, Greet, Jesse, Matthias), my colleagues at RoboRana (Maximilien, Arne, Lander), and my mentor Maarten.
On to the next step!
