Academic project · AI & Machine Learning
INFOCITY
I built INFOCITY as an academic project that connects government-service questions to document citations and escalates unresolved inquiries to staff.
My role: AI & Full Stack Engineer
The problem I faced
I needed to make scattered ordinances and advisory documents easier to search while giving citizens a path to staff when the evidence was insufficient.
My approach
I built INFOCITY as an academic project that connects government-service questions to document citations and escalates unresolved inquiries to staff.
Architecture decisions
Knowing When to Escalate
I apply a retrieval threshold before answering. When the source match is insufficient, I route the inquiry to city personnel. Similarity is a routing signal, not proof that an answer is correct.
What I delivered
- I built the interface with Next.js and TypeScript
- I implemented document ingestion with FastAPI
- I connected embeddings to semantic retrieval
- I routed unresolved questions to staff
- I added citizen feedback collection
Technology and methods
- Next.js
- TypeScript
- FastAPI
- Python
- ChromaDB
- LangChain
- OpenAI