TastePilot

Inspiration

Inspiration

I was inspired by a simple thought: finding something to do should be about personal taste, not just popularity.

When someone looks for a place to hang out with friends, they are not simply looking for a list of popular places. They want recommendations that match their interests, preferences, and plans.

I wanted to build an application that understands these preferences and helps people discover experiences that feel more relevant to them. This idea led us to build TastePilot using Qloo's cultural taste graph and AI.

Challenges We Faced

Building TastePilot came with a few interesting challenges:

• Understanding user requests: Converting natural-language prompts into structured information that could be mapped to Qloo's API.
• Finding relevant tags: Mapping user preferences to the tags supported by Qloo without losing the original intent.
• Improving recommendation quality: Retrieving results was not enough. We needed additional ranking and evaluation to identify more relevant recommendations.
• Building an agentic workflow: Designing a controller that could evaluate results, decide whether to retrieve more candidates, and determine when to stop.
• Deployment: Integrating multiple components, managing API credentials, and making the application publicly accessible.

Through these challenges, we learned how to combine LLMs, APIs, cultural intelligence, and agentic workflows to build a practical recommendation application.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for TastePilot

Built With

• agent
• python
• sqlite
• streamlit