CraveGuru

The engine behind "what should we do tonight?"

CraveGuru is a multi-category decision assistant built to solve a universal problem: too many options, too little signal. Instead of browsing endlessly, users set a few preferences — group size, budget, mood — answer a handful of targeted questions, and get a matched recommendation. Fast.

Four categories, one engine

The same recommendation engine powers four distinct use cases — each with its own question set and result surface, all sharing the same preference layer underneath.

• Date Ideas — activities matched to group size, budget, and vibe
• Steam Games — PC game picks based on mood, group, and price point
• Board Games — tabletop suggestions by player count and complexity
• Movie / TV Night — watch picks filtered by genre mood and group preference

Preferences first, then questions

Users start by locking in the context: how many people, price per person, and an optional mood tag from 18 vibe options (romantic, adventurous, cozy, spontaneous, etc.). Then a configurable 3–10 question sequence narrows the match. More questions = tighter result. The slider gives users control over precision vs. speed.

• Group size: 1–6 people
• Budget tiers: Free → $50+ per person
• Mood taxonomy: 18 vibe tags for personalization
• Question depth: adjustable 3–10 question slider ("More = better match")
• Group mode: shared session for multiple decision-makers

Serverless engine, three surfaces.

One recommendation engine powers a web app, a mobile app, and a marketing site — all sharing the same preference layer. The backend is AWS Lambda + Node.js; the web client is React + TypeScript; mobile is React Native + Expo; the landing site ships to S3 + CloudFront.

Live. Expanding.

Crave Guru is live and running. The core preference + question engine is stable across all four categories. The recommendation model continues to improve as usage data grows. Additional categories are planned.