Restaurant AI
Social restaurant discovery platform with visual collections, real-time group dining chat, and multi-agent personalization. 4 orchestrator-delegated specialist agents handle Yelp discovery, flavor profiling (6-dimensional taste vectors), beverage pairing, and budget analysis. Hybrid allergy filtering — keyword intersection confirmed by AI — runs as an independent safety gate before preference scoring across the full agent layer.
Picking a restaurant a whole group can eat at — safely — is a tangle of allergies, tastes, and budgets that most apps ignore. Potluck coordinates the group and screens for safety before it recommends. Full-stack social restaurant platform with a visual discovery feed. The feed lets users save restaurants to collections, follow friends, and explore trending spots by cuisine or taste profile. A real-time group chat layer coordinates dining decisions — the recommendation engine aggregates preferences across group members and surfaces options everyone can eat safely. Multi-agent orchestration (Strands) delegates to four specialists: Yelp discovery, flavor profiling (6D taste vectors: sweet/salty/sour/bitter/umami/spicy), menu-aware beverage pairing, and budget analysis. Hybrid allergy filtering runs keyword intersection then AI confirmation before any preference scoring. AWS Bedrock Guardrails enforce content safety across the full agent layer.
Restaurant discovery is either solo and algorithmic or group-based and uncoordinated. Allergy filtering in most apps is keyword-only and misses cross-contamination and ingredient aliases. Flavor preferences are flattened to categories instead of modeled as the multi-dimensional signal they are — and no app combines personalization, safety, and real-time group coordination in one place.
Visual discovery with taste-matched feeds. A group chat layer coordinates dining decisions in real time, with the recommendation engine aggregating preferences and surfacing safe options for everyone. Hybrid allergy filtering (keyword intersection + AI confirmation) runs as a safety gate before any preference scoring. Orchestrator-mediated specialists handle Yelp discovery, 6D flavor profiling, menu-aware pairing, and budget analysis.
