AI Graduation project idea

[Aly-EL-Badry]

Body

I am in my last year in Cairo University, and I face a problem with selecting my graduation project,
so can u help me by recommending project ideas that are not done.,,or can u drop the problems u face and hope to be solved by AI?
Thanks for ur help 🙏

Guidelines

• I have read and understood this category's guidelines before making this post.

[midiakiasat]

Hi Aly,

Strong graduation projects are defined by problem clarity, not novelty claims.

Here are AI project directions that are defensible, scoped, and evaluable at graduation level:

1. AI Failure Detection (not prediction)
   Detect when a model/system output becomes unreliable or unsafe.
   Focus: confidence collapse, boundary conditions, uncertainty.

2. Verifiability of AI Outputs
   Build a system that determines whether an AI answer is verifiable (traceable, falsifiable, consistent).
   Focus: trust, evidence, evaluation criteria — not accuracy alone.

3. Dataset / Concept Drift Detection
   Detect when real-world input diverges from training data over time.
   Focus: measurable drift signals and silent failure prevention.

4. Human-in-the-Loop Decision Boundaries
   Design AI that explicitly decides when not to decide and escalates to a human.
   Focus: thresholds, risk, responsibility.

5. Constraint Satisfaction AI (validation, not generation)
   Verify whether outputs satisfy strict rules (legal, medical, engineering).
   Focus: rule enforcement, edge cases, explainability.

Selection rule (simple and strict):
If you can clearly state failure mode, dataset, metric, and success condition, the project is valid.

Avoid hype. Optimize for clarity, rigor, and evaluation.

[kmoragap]

Don't obsess over finding an idea that's "never been done," you will just get stuck. It is better to take a working concept and apply it to a niche you actually care about.

Here are some repos to check for inspiration, the last one is best for AI:

https://github.com/topics/final-year-project

https://github.com/Projects-Developer/50-Final-Year-Projects-with-Source-Code

https://github.com/DharminJoshi/Final-Year-Project-Ideas-2025

https://github.com/youssefHosni/Awesome-AI-Data-Guided-Projects

Since you are in Cairo, lean into that. Generic AI tools are everywhere, but there are huge gaps in things like Egyptian dialect NLP or computer vision for local traffic. Professors usually appreciate a solved local problem way more than a generic copy-paste app.

Focus on a specific angle rather than a massive platform.

Good luck!! :)

[healer0805]

"1. AI that detects wrong AI answers"
Build a system that flags hallucinations or contradictions instead of generating answers.

"2. Explainable AI decisions"
AI gives an answer "and" a clear reason based on rules (education, finance, or law).

"3. Workflow mistake detector"
AI checks if steps are missing or done in the wrong order (forms, labs, coding).

"4. Offline / low-resource AI"
AI that works without internet or GPU (education or health use case).

"5. Instruction → checklist AI"
Turns messy human instructions into clean, ordered steps.