Computer Science & Software · Updated 2026-10-14
Final Year Project Ideas for Computer Science and Software Engineering (2026).
When evaluating impactful final year project ideas in Pakistan for CS and software engineering degrees, combining AI models or cloud microservices with a demonstrable fyp prototype significantly outperforms pure screen-only applications during external viva examinations.
40 software-first project ideas in AI, generative AI, web and mobile, security and systems, with stack, difficulty and data source for each.
Here are 40 final year project ideas for computer science and software engineering students in Pakistan, grouped by area. Each idea lists a suggested stack, a difficulty level and where to get data. Software projects score best when they solve a local problem, use real data and report clear evaluation numbers, so each idea is written with that in mind.
AI and machine learning FYP ideas
| Idea | Suggested stack | Difficulty | Data source |
|---|---|---|---|
| Crop disease detection app that works offline | TensorFlow Lite or ONNX, Android | Medium | Public plant disease datasets plus your own field photos |
| Urdu handwritten text recognition (OCR) | CNN plus sequence model, Python | Hard | Public Urdu handwriting datasets; collect your own samples |
| Pakistan Sign Language recognition | MediaPipe pose, LSTM or transformer | Hard | Record signs with volunteers; check existing PSL datasets |
| Fraud detection for mobile wallet transactions | XGBoost, anomaly detection, dashboard | Medium | Public synthetic financial transaction datasets |
| Student dropout early-warning system | Scikit-learn, web dashboard | Medium | Anonymised records with your university's permission |
| Traffic density estimation from CCTV footage | YOLO, OpenCV, simulation | Hard | Recorded junction videos (get permission) |
| Electricity demand forecasting for a housing society | Time-series models, Python | Medium | Smart-meter or bill data from volunteers |
| Face-recognition attendance with liveness detection | Face embeddings, anti-spoofing model | Medium | Consenting classmates; follow privacy rules |
| Medical image triage assistant (research demo, not for diagnosis) | CNN, explainability heat-maps | Hard | Public chest X-ray or skin-lesion datasets |
| Property price estimator for Lahore, Karachi or Multan | Gradient boosting, map front end | Medium | Public listings, collected within each site's terms |
Generative AI and AI agent FYP ideas
Generative AI projects are popular, so examiners look harder at evaluation. Build a test set of real questions or tasks and report how often the system gets them right.
| Idea | Suggested stack | Difficulty | Data source |
|---|---|---|---|
| University help-desk chatbot that answers from official documents (RAG) | LLM API or open model, vector database, web app | Medium | Prospectus, fee notices, academic rules |
| Urdu and Roman Urdu customer-support assistant for a local business | Multilingual LLM, RAG, WhatsApp-style chat interface | Medium | Business FAQs and past (anonymised) chats |
| AI agent that completes multi-step admin tasks in a test web portal | LLM with tool calling, browser automation | Hard | A mock portal you build for testing |
| Legal document explainer for tenancy or sale agreements in Urdu | LLM, document parsing, citations to clauses | Medium | Sample agreements (no personal data) |
| Exam question generator and auto-grader for a course | LLM, rubric-based grading, teacher review screen | Medium | Past papers with your instructor's permission |
| Code review assistant for student programming labs | LLM, static analysis, Git integration | Medium | Anonymised lab submissions |
| Meeting minutes generator for Urdu-English meetings | Speech-to-text, summarisation model | Medium | Recorded mock meetings |
| Voice assistant for farmers in Punjabi or Saraiki | Speech recognition, text-to-speech, weather and price APIs | Hard | Recorded voice samples from volunteers |
Web and mobile app FYP ideas
| Idea | Suggested stack | Difficulty | What to evaluate |
|---|---|---|---|
| Blood donor matching app with location search | Flutter or React Native, maps API | Easy | Time to find a matching donor |
| Point-of-sale and e-invoicing app for small shops | Web or mobile app, offline-first database | Medium | Time per sale, stock accuracy |
| Home tutor marketplace with booking and reviews | Full-stack web app, payment sandbox | Medium | Usability testing with real users |
| Hostel and mess management system | Web app, role-based access | Easy | Admin time saved per week |
| Public transport tracker for a city bus route | GPS app, real-time map | Medium | Arrival time prediction error |
| Complaint management portal for a housing society | Web app, ticketing, SMS alerts | Easy | Time to resolve complaints |
| Waste collection route planner | Route optimisation, mobile app for drivers | Medium | Distance saved vs current routes |
| Accessible app for visually impaired users | On-device object detection, text-to-speech | Medium | Task success in user trials |
Cybersecurity FYP ideas
| Idea | Suggested stack | Difficulty | What to evaluate |
|---|---|---|---|
| Phishing SMS and WhatsApp message detector | Text classifier, Android app | Medium | Precision and recall on labelled messages |
| Network intrusion detection system with a honeypot | Zeek or Suricata, machine learning, dashboard | Hard | Attacks detected in a lab network |
| Secure file sharing with end-to-end encryption | Web app, public-key cryptography | Medium | Security review and performance tests |
| Password manager with breach checking | Browser extension, local encryption | Medium | Threat model and penetration test results |
| Deepfake image detector | CNN, explainability | Hard | Accuracy on held-out fake images |
| Vulnerability scanner for small business websites (with owner permission) | Python, OWASP checks, report generator | Medium | Issues found on test sites you control |
Data, cloud and systems FYP ideas
| Idea | Suggested stack | Difficulty | What to evaluate |
|---|---|---|---|
| University timetable optimiser | Constraint solver (e.g. OR-Tools), web front end | Medium | Clashes removed, solve time |
| Load-shedding schedule and outage alert service | Data scraping (within terms), notifications, API | Medium | Alert accuracy and delay |
| Microservices e-commerce platform with auto-scaling | Docker, Kubernetes, monitoring | Hard | Throughput and recovery under load tests |
| Open data portal for district-level statistics | Data pipeline, charts, search | Medium | Data freshness, user testing |
Software projects with a hardware component
Adding a small physical device can make a software project easier to demonstrate, but it should support the software, not replace it. Keep the hardware simple and spend your effort on the app, the data and the evaluation.
| Idea | Suggested stack | Difficulty | What to evaluate |
|---|---|---|---|
| Smart energy dashboard with an ESP32 meter | ESP32, MQTT, web dashboard | Medium | Reading accuracy, dashboard response time |
| Real-time defect detection on a small conveyor | Camera, edge AI model, operator app | Hard | Detection accuracy at line speed |
| Offline Urdu voice commands for home appliances | On-device keyword spotting, relay board | Hard | Command accuracy in a noisy room |
| Drone imagery crop health map (images captured by a licensed operator) | Image stitching, vegetation index, web map | Hard | Agreement with field observations |
For apps that control physical devices, see final year app projects with hardware and IoT. If you need a sensor box, PCB or enclosure built, PakMEC's FYP development support can fabricate it so you can focus on the software.
What examiners look for in a CS or SE final year project
- A clear problem and user: who uses the system, and what they do today without it.
- Requirements and design documents: use cases, architecture diagram, database design and the reasons behind your technology choices.
- Evaluation with numbers: model metrics on a held-out test set, load-test results or user-testing scores. Compare against a simple baseline.
- Working deployment: a hosted demo or installable app, not only code on a laptop.
- Honest limitations: what fails, why, and what you would do next.
A two-semester plan for software FYPs
Semester 1: problem, data and prototype
Finish the proposal, collect or source the data, build a first working version of the core feature and run a baseline evaluation before the mid-term review.
Semester 2: build, test and deploy
Complete the remaining features, improve the model or system against your baseline, run user testing, deploy it, and write the report as you go rather than in the final month. The proposal guide and testing guide cover both stages.
Looking for ideas in other departments? See the full list of 64 FYP ideas.
Frequently asked questions
Is a chatbot a good FYP for computer science?
It can be, if it answers from a defined set of documents, works in Urdu or Roman Urdu, and you measure its answer accuracy on real questions. A thin wrapper around a public chatbot with no evaluation will not satisfy most panels.
Can computer science students do hardware-based projects?
Yes, if the software is the main contribution. Use ready-made modules for the hardware and put most of your effort into the application, data processing and evaluation.
How big should a dataset be for an AI FYP?
There is no fixed number. Start with a pre-trained model and fine-tune it on a few hundred to a few thousand labelled local examples, then show how accuracy changes as you add data.
Can I use paid AI APIs in my FYP?
Yes, but budget for the cost and keep a fallback, such as a smaller open model, in case your credits run out before the defence.
Need help with the next step?
Need Hardware or Prototyping Support for Your CS/SE Final Year Project?
PakMEC helps software and computing students across Pakistan integrate real-world hardware, custom PCBs, edge computing boards, and viva-ready documentation.