Website-Deployed Engineering Knowledge Assistant
A retrieval-augmented assistant for exploring project materials, technical notes, and engineering knowledge through grounded question answering.
Role
Applied AI Engineer
Focus
Applied AI · RAG · Knowledge Systems
Status
In Progress
Overview
This project turns a personal portfolio into an interactive knowledge surface. Instead of just listing projects, it makes technical work explorable through grounded retrieval and citation-aware responses.
Problem / Context
Engineering work often lives across notes, project docs, architecture decisions, and code-adjacent artifacts that are hard to browse linearly.
I wanted an interface that could help recruiters, collaborators, or technical peers ask targeted questions and get grounded answers quickly.
What I Built
Designed a website-integrated RAG concept for project materials, engineering notes, and personal technical knowledge.
Focused on metadata-aware chunking and citation-backed retrieval so answers remain grounded and easier to trust.
Evaluated the system around relevance, groundedness, and latency rather than treating it as a purely aesthetic AI feature.
Technologies
RAG, Embeddings, Vector Retrieval, Metadata Chunking, Evaluation
Key Engineering Decisions
The interface is designed as a product surface, not just a model demo, so interaction quality and trust signals matter as much as retrieval quality.
I kept the content model structured so future ingestion and assistant features can grow without redesigning the whole website.
Outcome / Significance
The project creates a natural future extension point for a portfolio that is both expressive and technically legible.
It reflects how I think about applied AI: grounded, useful, and integrated into real user-facing systems.
Future Work
Add document ingestion and citation rendering.
Introduce conversation memory and richer evaluation tooling for user-facing queries.