Yes, if you choose the right kind of AI project. Applied AI projects that build a useful system around an existing model, such as retrieval assistants, text-to-SQL tools or document processing, need solid engineering far more than advanced maths. Avoid projects whose main contribution is a new model or training method.
What people tell me
Our final year project groups are forming and most people want to do something in AI. I find AI interesting, but I struggled with linear algebra and statistics and I am worried I will choose an AI project, get stuck on the maths and fail to deliver. At the same time I feel a non-AI project will look outdated to employers. Should I still do an AI final year project?
A composite of the messages behind this question, with personal details left out.
Key takeaways
- Applied AI projects need engineering, evaluation and product sense more than maths.
- Avoid projects whose core contribution is training or inventing a model.
- Scope around one real user and a measurable result, not a vague chatbot.
- Plan evaluation from the start; it is what separates good AI projects.
- A strong non-AI project still beats a weak AI one with employers.
Two very different kinds of AI project
When people say "AI project" they usually mean one of two things, and the maths needed is completely different.
| Model-centred project | Applied AI project | |
|---|---|---|
| Main contribution | A new or trained model, a better accuracy score | A useful system built around existing models |
| Typical work | Training, tuning, architecture changes, experiments | Retrieval, prompts, tools, data pipelines, UI, evaluation |
| Maths needed | Linear algebra, probability, optimisation, often a lot | Basic statistics for evaluation, little else |
| Risk if maths is weak | High: you get stuck and cannot explain results | Low: the hard parts are engineering |
| Jobs it points to | ML research, ML engineering | AI engineering, backend, full stack |
If maths is not your strength, the applied side is where you can do strong work. It is also where most of the current job demand sits.
What a good applied AI project looks like
A good one solves a real problem for a specific group of people and measures whether it worked. For example:
- A retrieval assistant for university regulations that answers student questions with citations to the exact rule.
- A guarded text-to-SQL assistant that lets a small business owner ask questions about their sales data safely.
- A claim verification assistant that checks statements against sources. This one is more advanced, so only take it on as a strong group.
All three use existing models. The difficult parts are retrieval quality, handling wrong answers, security, user experience and evaluation. Those are engineering problems.
The trap: a chatbot with no evaluation
Many AI final year projects are a chat interface over one model call, demoed with three hand-picked questions. Examiners and employers have seen hundreds of these. What makes yours different is evidence:
- A test set of 50 to 100 real questions with expected answers.
- A measurement of how often the system is right, and how it fails.
- One or two improvements you made and the numbers before and after.
That is basic statistics, not advanced maths. You can do it.
When to avoid AI anyway
Choose a non-AI project if your group has no one comfortable with backend work and data, if your supervisor wants a model-centred contribution you cannot deliver, or if you have a genuinely better idea elsewhere. A well-built system in web, mobile, data or security with real users is still stronger than a weak AI demo. Browse the FYP ideas library across domains before deciding.
How much maths you should still learn
Enough to understand what embeddings and similarity mean, what precision and recall measure, and why a model can be confidently wrong. That is a few weeks of focused effort, not a semester. I wrote more about this in Can I have a career in software or AI if I am bad at math?.
Where to learn what the project needs
The AI engineering roadmap covers LLM APIs, retrieval, tools and evaluation in order, and the AI engineering cheat sheet is a compact reference for the terms. Building Effective Agents and RAG Techniques are free and practical.
If you are still stuck
Read Which final year project will actually help me get a job? and Which AI topics are worth learning?. If you want to talk through your project idea and scope, join Sefism and, as a member, book a 1:1 session.
Was this answer helpful?
Read next
- An AI Engineering Roadmap That Puts Things in the Right OrderAn AI engineering roadmap in seven stages, from tokens and prompting through RAG, evals, tool calling, MCP and agents, with a project to build at each step, what to skip, and a realistic timeline.
- RAG, Fine Tuning, or a Better Prompt: How to ChooseRAG vs fine tuning vs a better prompt: a practical order from prompting through retrieval, embeddings, reranking, tools and agents to fine tuning, with the signal that tells you to move to the next step.
Your situation is not quite this one?
Members get written answers to their own questions, a roadmap built for them and feedback on their projects. Early access is open: sign up and your account is active shortly after. Prefer to talk? A free call works too.