Product & AI
Unifeast
A multi-turn dining recommendation chatbot that uses retrieval and rule-based allergen filtering outside the language model.
- Role
- AI/ML Engineer, Amazon University Engagement Programme
- Project type
- Product & AI
- Status
- Prototype
What it is
Unifeast is a dining recommendation chatbot built during the Amazon University Engagement Programme. It helps users navigate recommendations across multiple turns while treating allergen safety as a system constraint rather than a generated suggestion.
Contribution
As sole developer, I built the backend through three iterations, moving from the Bedrock console to LangChain. The final design keeps a rule-based allergen filter outside the language model. The project received Best Technical Implementation.
Case status
The project currently has a four-tab mobile demo, sample data, a small RAG implementation, and an evaluation record. The full case will document these artefacts without overstating the scope of the prototype.