20 BSc CSIT Final Year Project Ideas (With Tech Stack & Scope)
20 BSc CSIT final year project ideas with tech stack, main modules, algorithms and difficulty, from RFID attendance to IoT dashboards.
On this page
Save this list before your group meeting. Choosing a title is the hardest part of the project: pick something too small and it looks thin, pick something too large and it never finishes. Here are 20 ideas with a stack, the main modules and the algorithm you can point to, so your proposal feels solid from day one.
A common piece of advice among students and supervisors is that a CSIT project is easier to defend when it shows at least one clear algorithm or technique, not only forms and database tables. This is general advice, not an official rule, so always check your department's format and your supervisor's expectations. Difficulty here assumes a team of two to four people.
Part A: Web and mobile systems
1. Smart attendance system with face recognition. Stack: Python, OpenCV, Flask, MySQL. Modules: enrolment, live recognition, reports, admin. Algorithm: face embeddings with cosine distance, or LBPH. Level: Medium.
2. Local-language news classifier. Stack: Python, scikit-learn, Django. Modules: scraper, cleaning, classifier, web UI. Algorithm: TF-IDF with Naive Bayes or SVM. Level: Medium. Nepal angle: Nepali text needs its own stop-word list and tokeniser.
3. Job and internship matching portal. Stack: Laravel or Django, PostgreSQL. Modules: profiles, job posts, matching, notifications. Algorithm: skill-overlap scoring (Jaccard similarity) with ranking. Level: Medium.
4. Bus route and fare finder. Stack: React, Node.js, MongoDB. Modules: stops and routes, search, admin panel. Algorithm: Dijkstra or A* shortest path. Level: Medium. Nepal angle: microbus routes in the valley are poorly mapped online.
5. Crop price tracker for farmers. Stack: Flutter, Firebase or Laravel API. Modules: price entry, market comparison, alerts. Algorithm: moving-average trend prediction. Level: Easy.
6. E-library with smart recommendations. Stack: Spring Boot or Django. Modules: catalogue, borrowing, recommendations. Algorithm: collaborative filtering or content-based similarity. Level: Medium.
7. Online exam system with cheating detection. Stack: MERN. Modules: question bank, timer, randomisation, tab-switch logging, results. Algorithm: question shuffling with seeded randomness, anomaly rules. Level: Medium.
8. Hospital appointment and queue manager. Stack: Laravel, MySQL. Modules: doctors, slots, tokens, SMS reminders. Algorithm: priority queue scheduling. Level: Easy.
Part B: Data, AI and security
9. Fake-review or spam message detector. Stack: Python, scikit-learn. Modules: dataset, features, model, API. Algorithm: logistic regression versus SVM, compared with precision and recall. Level: Medium.
10. Student performance predictor. Stack: Python, pandas, Streamlit. Modules: data upload, model, dashboard. Algorithm: decision tree or random forest. Level: Easy. Tip: use anonymised or self-collected data and mention ethics.
11. Plant disease detection app. Stack: Python, TensorFlow Lite, Flutter. Modules: image capture, model, advice screen. Algorithm: CNN with transfer learning (MobileNet). Level: Hard.
12. Nepali handwritten digit recogniser. Stack: Python, Keras. Modules: dataset creation, training, web demo. Algorithm: CNN, with a comparison against k-NN. Level: Medium.
13. Password strength and breach checker. Stack: Node.js or Python. Modules: entropy scoring, rule checks, hashed lookup. Algorithm: strength estimation and privacy-preserving breach lookup. If the app stores login passwords, use Argon2id or bcrypt. Level: Easy.
14. Network intrusion detection dashboard. Stack: Python, Scapy, a public dataset such as NSL-KDD. Modules: capture or dataset loader, model, alerts. Algorithm: random forest classifier. Level: Hard. Only capture traffic on networks you own or have permission to monitor.
15. Image steganography tool. Stack: Python or Java. Modules: embed, extract, key handling. Algorithm: LSB embedding with AES encryption. Level: Medium.
Part C: Software plus hardware
16. RFID attendance with web dashboard. Stack: ESP32, RC522, MQTT or HTTP, Laravel, MySQL. Modules: card enrolment, scanning device, late-arrival rules, reports. Algorithm: time-window rules and duplicate-scan filtering. Level: Easy to Medium. This is a good first hardware-integrated project.
17. IoT environment dashboard. Stack: ESP32, DHT22, MQTT, Node-RED or React, InfluxDB. Modules: device firmware, broker, storage, charts, alerts. Algorithm: moving average and threshold alerts. Level: Medium.
18. Smart parking slot finder. Stack: ESP32 with ultrasonic or IR sensors, Firebase, Flutter. Modules: sensing, cloud sync, map screen. Algorithm: nearest-free-slot search. Level: Medium. Nepal angle: parking pressure in Kathmandu and Pokhara markets.
19. Smart agriculture advisor. Stack: ESP32, soil moisture, DHT22, Python rule engine, web app. Modules: sensing, irrigation advice, history. Algorithm: rule-based expert system, optionally a decision tree. Level: Medium.
20. Load-shedding aware appliance scheduler. Stack: ESP32, relay (low-voltage demo), mobile app. Modules: schedule input, supply detection, auto start. Algorithm: interval scheduling. Level: Medium. Safety: demonstrate with low-voltage lamps; mains wiring only by a qualified electrician.
Quick comparison table
Scroll sideways to see all columns.
| Idea type | Hardware needed | Algorithm clarity | Risk |
|---|---|---|---|
| Web portals (3, 6, 8) | None | Medium | Looks like a plain CRUD app |
| ML projects (2, 9, 10, 11, 12) | None | High | Dataset quality |
| Security (13, 14, 15) | None | High | Scope creep |
| IoT (16 to 20) | Yes | Medium | Parts and wiring delays |
How to make any idea stronger
- State the problem in one sentence and say who has it.
- Name your algorithm and explain why you chose it over one alternative.
- Measure something. Accuracy, response time, number of users supported, battery life.
- Build a baseline first, then improve it.
- Write down limits. What does the system not do?
Scope check: is your idea the right size?
Run your idea through this short test before you submit the title.
Scroll sideways to see all columns.
| Check | Healthy answer |
|---|---|
| Can you list five to eight modules? | Yes, and each has an owner |
| Is there one thing you can measure? | Accuracy, speed, users or uptime |
| Do you have the data or parts? | Yes, or a plan to collect them in two weeks |
| Can a basic version work in a month? | Yes |
| Can you explain the algorithm on paper? | Yes, with a small worked example |
If the answer to two or more is "no", shrink the idea. A smaller project with real results and clean documentation is easier to defend than a big one that is only partly working.
What to avoid
- Copying a GitHub project and changing the logo. Examiners ask about the code.
- Using a dataset you cannot explain.
- Promising mobile app, web app, AI and IoT in one semester.
- Leaving hardware wiring to the last week.
FAQ
Do I need an algorithm in my CSIT project?
Many students and supervisors say it helps in the viva because it gives you something specific to explain. It is common advice, not a universal rule, so confirm with your department.
Which idea is easiest to finish on time?
The attendance, appointment or crop price projects have clear scope and a plain database. Add one extra feature rather than five.
Can I mix hardware with a software project?
Yes. An ESP32 sending data to a web dashboard is a popular combination, but plan parts early and keep the hardware simple.
Which language should I choose?
Use the one your team already knows. Python suits ML, JavaScript or PHP suits web, and Dart suits mobile.