The challenge
The problem worth solving.
Project and grant proposals are frequently assessed against complex criteria, yet applicants may receive limited, late or inconsistent feedback before submission.
Master’s Research & Engineering
My MSc dissertation translated a complex research question into a working AI-assisted platform that combined human-centred design, secure deployment and evidence-aware assessment.
The challenge
Project and grant proposals are frequently assessed against complex criteria, yet applicants may receive limited, late or inconsistent feedback before submission.
The approach
I used a Design Science Research approach to build and evaluate a secure platform with a React interface, FastAPI services, PostgreSQL persistence and locally hosted language models. The design explored structured prompting, document parsing and retrieval techniques for grounded proposal feedback.
My contribution
Designed the research methodology and translated it into a working technical artefact.
Built the React frontend, FastAPI backend and PostgreSQL data layer.
Integrated locally hosted LLMs to support privacy-conscious processing.
Explored hybrid parsing, graph-informed retrieval and causal reasoning concepts.
Experience it here
Live previews, walkthroughs and media remain part of the case study so visitors can explore without losing their place in Sam Xperience.
Embedded preview
Screens, walkthroughs or a live product environment will appear here without taking visitors away from Sam Xperience.
What it taught me
“AI assistance is most valuable when its limits are visible, evidence remains traceable and the human user retains ownership of the final judgement.”
Continue the exhibition