TAVA: Taste-Aware Venture Architect

Inspiration

The inspiration for TAVA came from witnessing the gap between entrepreneurial dreams and data-driven execution. Entrepreneurs would spend months developing business idea only to discover there was no viable market, or they'd enter saturated markets without understanding their competitive landscape. The idea was to provide with data-driven approaches with market intelligence, making it accessible to creators and entrepreneurs at the earliest stages of their journey.

What it does

Tell TAVA what kind of place you want to open, where you'd like to open it, and what inspires you.
TAVA analyzes trends, audience signals, and market data using Qloo and Gemini.
Explore audience segments, trends, and an interactive heatmap for your chosen location.
Refine your concept and location to develop the best strategy.
Receive a tailored business concept, SWOT/ESG analysis, and a downloadable report.

How I built it

It is built on a Flask backend and leverages multi-step reasoning with Gemini Pro. The app integrates the Qloo API and Google Places API, and is designed to be stateless, processing all data through real-time API calls.

Challenges I ran into

I initially struggled with understanding Qloo’s unique concepts and how to effectively apply its API. Integrating its cultural insights with Gemini’s reasoning pipeline also required careful data handling and architecture design.

Accomplishments that I'm proud of

I’m proud of creating a fully functional prototype that combines cultural intelligence with LLM to deliver real-time, personalized business strategy. Successfully integrating multiple APIs into a seamless user experience was a key milestone.

What I learned

I learned how to work with taste-based recommendation systems like Qloo, and how to design a prompt pipeline that blends structured data with generative reasoning. Building the web interface with Flask and deploying it on Google Cloud Run was a fun and rewarding process.

What's next for TAVA

Next, I plan to improve location-specific accuracy, add more visual insights like competitive mapping, and allow users to export branded pitch decks directly from the platform. I'd also like to explore more real-time feedback loops using user preferences.