Beyond Vibe Evaluations: Evaluating AI

Agenda:

5-Min Community Talk
Time: 5.15pm-5.20pm

Title: Agent, Plan me a Holiday!
I want to talk about my project WanderAwai, which is an AI-powered travel planning web app designed to simplify trip planning by offering personalised itineraries and smart recommendations. Using Azure OpenAI, it generates custom travel plans based on user preferences, budgets and dates. The app suggests flights, hotels, restaurants, and activities with plans to integrate live APIs for real-time data. WanderAwai also aims to foster a travel community where users can share tips and experiences. Built on Microsoft Azure with a modern tech stack including Next.js and Node.js, it showcases how AI and cloud technology can create seamless, personalized, and social travel planning experiences.
The key takeaway will be to see how AI-driven tools and real-time data can create seamless, custom itineraries that simplify the user experience.

Speaker: Mihir Prashant Tayshete - University Student (Postgraduate/Master's)
I am a passionate technologist recently completed my Master’s in Information Technology at UWA. I’m deeply interested in data-driven solutions, AI and full-stack app development. I enjoy turning complex challenges into scalable, intuitive applications.

Time: 5.25pm-6.10pm
Title: Beyond Vibe Evaluations: Evaluating AI
Did that prompt change actually lead to a better outcome?
Does the new data added break expected behaviours?
Our AI said what to a user?!?
To build AI systems you need more than intuition, you need evidence. This talk covers practical ways to evaluate AI solutions beyond the ‘vibe check’ and explores how to move beyond guesswork and build reliable, measurable, and maintainable AI.