Inspiration
As Gen Z creators and dreamers, we often find ourselves searching for experiences that feel like us — not just algorithmically “correct.” The world is overflowing with recommendations, but few of them understand our unique vibes, passions, or cultural quirks. We wanted to build something that listens deeply, understands your taste, and curates experiences that feel both magical and personal. Thus, TasteMuse was born — an AI cultural concierge that crafts personalized lifestyle and travel plans using the power of LLMs + Qloo's Taste AI™.
What it does
TasteMuse takes a simple input like: “I love Wes Anderson films, indie bookstores, Ethiopian jazz, and low-key coffee shops.” And outputs a tailored lifestyle plan — your perfect day, curated with:
Cultural and aesthetic travel suggestions Local food/dining spots that match your vibe Music and movie recs aligned with your tastes Locations pulled from Qloo's global insights All wrapped in natural, human-like storytelling from GPT-4o It’s like having a creative best friend, an intuitive trip planner, and a cultural critic — all in one.
How we built it
We combined: Qloo’s Taste AI™ API to interpret user input and extract culturally relevant insights (e.g., food, music, fashion, travel preferences) OpenAI GPT-4o to craft a highly personalized narrative — a day-in-the-life experience FastAPI backend to orchestrate the API calls Our prompt design was critical — we injected Qloo's data into the GPT prompts in a way that maintained emotional, human-first storytelling.
Challenges we ran into
LLM prompt tuning: Balancing structure with creativity to avoid robotic or repetitive responses Qloo API query design: Understanding how to extract structured insights (like cuisine types or musical genres) from freeform taste input Keeping the soul: It’s easy to build another recommendation engine — it’s hard to make it feel like art.
Accomplishments that we're proud of
Building an emotionally intelligent AI assistant in <48 hours Using Qloo’s API in a genuinely expressive, non-transactional way Creating something that feels human-first — not just smart, but soulful Testing successfully with diverse user inputs (music lovers, cinephiles, foodies)
What we learned
Taste is a fingerprint — no two users want the same "top 10 list" Integrating a cultural knowledge graph (Qloo) with an expressive LLM yields much richer experiences than either tool alone. Prompt design is an art form — and it evolves with every user interaction
What's next for TasteMuse
Build a full mobile-first interface Add memory so TasteMuse remembers your evolving tastes Geo-personalized suggestions based on real-time city data Collaborate with tourism boards or lifestyle brands for hyper-targeted cultural plans A/B test tones (sarcastic, poetic, minimalist, etc.) to match personality preferences We believe the future of AI is not just smart — it's human, weird, and wonderfully emotional.
Log in or sign up for Devpost to join the conversation.