Retrieval Assistant for University Regulations
A question answering system over your own university's handbook, examination rules and fee policies that answers in plain language and cites the exact clause it used. The interesting engineering is not the model, it is refusing to answer when the documents do not cover the question.
Intermediate
A small team, or one strong student willing to learn something new.1 semester, 2 students
5 to ship
3 optional extrasSuggested stack
What you should ship
- Ingestion pipeline that chunks the source PDFs and stores embeddings with page and section references
- Retrieval and answer service that returns an answer plus the passages it was drawn from
- Web interface showing the answer with each citation linked to the source page
- A test set of at least 100 real student questions with expected answers, and measured accuracy against it
- Refusal behaviour: questions outside the corpus return a clear 'not covered' response rather than a guess
If you have time left
- Handle document updates by versioning chunks so old answers can be traced to the policy in force at the time
- Support a second language alongside English
- Confidence score shown to the user, calibrated against the test set
The problem
Every university has a handbook nobody reads, a regulations document nobody can navigate, and an office that answers the same forty questions all year. The information is public and correct; it is simply unfindable.
What you build
An ingestion pipeline that splits the official documents into passages and embeds them, a retrieval layer that finds the relevant passages for a question, and an answer service that produces a plain language response grounded strictly in those passages, with citations back to the source page.
Why it works as a final year project
The scope is naturally bounded by a document set you actually have, the evaluation is objective because the correct answer is written down somewhere, and the failure modes are interesting rather than trivial.
The hard part
Refusal. Any system like this will confidently answer questions the documents do not cover, and that is the failure that matters, because a wrong answer about examination eligibility is worse than no answer. Building a system that reliably says it does not know is the genuine contribution here, and it is what your evaluation should focus on.
How to evaluate it
Build the question set early, before you build the system. A hundred real questions collected from students, each with the correct answer and the clause that supports it, plus twenty questions that the documents deliberately do not answer. Report retrieval accuracy, answer accuracy and refusal rate separately.
Scope warning
Do not train or fine tune anything. Use an existing model through an API or a small local one. The project is the retrieval, the grounding and the evaluation, and every hour spent on model training is an hour not spent on the part that is actually being assessed.
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
Not sure this one fits you?
I help students pick an idea that matches their skills and their deadline, then scope it down until it is finishable. Guidance only, never done for you.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Every session is free; a few slots open each week.
Follow along
New writing, resources and project ideas land here first.