A RAG-powered assistant prototype for construction-domain questions.
Led the team as captain and built a Retrieval-Augmented Generation kiosk assistant for the AI Future Minds'25 Construction AI track, using Hugging Face models to process construction-domain material and answer with grounded context.
View
Pattern
RAG
Role
Captain
Track
Construction AI
Role
Team captain / AI prototype
Context
AI Future Minds'25 Construction AI
Outcome
Applied LLM system
Engineering highlights
Led the team as captain while keeping the prototype scope focused under hackathon pressure.
Designed the assistant around retrieval and grounded responses instead of a generic chatbot flow.
Used Hugging Face tooling to connect model behavior with construction-specific source material.
Moved from model fascination to a concrete assistant pattern with retrieval, context, and domain constraints.
The kiosk format made the product requirement sharper: answers needed to feel direct, trustworthy, and usable by someone standing in front of a terminal.
The core engineering lesson was that RAG quality depends as much on retrieval shape and content boundaries as it does on the model.
This project clarified the direction I want to keep building toward: AI systems that are useful because they are grounded in a real domain.
