All work

DOCUMENT-GROUNDED ENGINEERING AI / HONOURS THESIS / RESEARCH

Active research

Document-Grounded Engineering AI

An answer should show its working.

Research into engineering assistants that answer from source documents and make their evidence traceable.

System overview: Source documents to Chunks + retrieval to Grounded answers to Citation evaluation.
System overview of the thesis research and evaluation path.View full image ↗

MY CONTRIBUTION

Honours researcher · UQ

HIGHLIGHT

Research and evaluation design · 2026 to 2027

TOOLS & METHODS

Retrieval-augmented generation · Local embeddings · Evaluation · Engineering documents

01 / THE PROBLEM

The problem.

Engineering projects contain requirements, deadlines and decisions across many documents. An assistant needs to find the right evidence and remain grounded in it.

02 / MY CONTRIBUTION

What I worked on.

  1. 01

    Defined the research around citation-grounded question answering and requirement extraction.

  2. 02

    Organised the document corpus and designed an evaluation of citation accuracy and unsupported claims.

  3. 03

    Explored how grounded planning could support engineering work, supervised by Dr Jingyu Ge.

03 / PROGRESS & RESULTS

Where it stands.

The thesis is in the research and evaluation-design stage. I’m developing ways to check citation accuracy and identify answers that the source documents don’t support.

NEXT PROJECT

Sharpe Sauce / AlgoJam