Computer Engineering Project Topics for Final Year Students

What makes a final year topic hold up

A good project has one question you can answer with evidence. "An online shopping website" has no such question, because nothing about it can be measured. "Does rotating a QR code every ten seconds stop proxy attendance, and what does it cost in failed scans?" gives you something to build, test and report.

Before you commit, try to write one sentence naming the user, the problem and the number that will show it worked. If you cannot, the topic is still too broad.

Security and privacy topics

1. Client-side phishing URL detection extension — JavaScript, TensorFlow.js, Python, public phishing URL feeds
   Sending every visited URL to a server is itself a privacy problem. Train a small model on lexical URL features, run it inside the browser, and compare accuracy and latency with a server-side version.
2. Secret scanner for Git repositories — Python, Git hooks, GitHub Actions
   Developers often commit API keys by accident. Combine regular expressions with entropy checks in a pre-commit hook, then measure false positives on open source repositories.
3. Flow-based intrusion detection tested across days — Python, scikit-learn, XGBoost, CICFlowMeter
   Many IDS papers report inflated scores because of random train and test splits. Train on some days of CIC-IDS2017 traffic, test on other days, and report how much performance really drops.
4. End-to-end encrypted file sharing portal — Web Crypto API, Node.js or PHP, MySQL
   College and office portals usually store documents in readable form. Encrypt in the browser so the server only holds ciphertext, and design how keys are shared and revoked when a member leaves.
5. Cloud storage misconfiguration auditor — Python, boto3, MinIO
   Publicly exposed buckets are a common cause of data leaks. Build a tool that reads bucket policies, flags risky permissions and suggests a corrected policy, testing against a local S3-compatible server.

Distributed systems and DevOps topics

1. Replicated key-value store using Raft — Go or Java, gRPC, Docker Compose
   Consensus is easy to describe and hard to implement. Build leader election and log replication, kill nodes during writes, and show that committed data survives while recording the latency cost.
2. API gateway with adaptive rate limiting — Node.js or Go, Redis, k6
   Fixed limits punish normal users during spikes and still miss abuse. Compare token bucket and sliding window limits with one that adapts to backend load, using generated traffic.
3. Flaky test detector for CI pipelines — Python, GitHub Actions API, PostgreSQL, Grafana
   Tests that fail at random waste developer time and erode trust. Collect CI run history, score each test for instability, and quarantine the worst offenders automatically.
4. Kubernetes autoscaling on custom metrics — Kubernetes, Prometheus, KEDA, RabbitMQ
   CPU is a poor scaling signal for queue-driven workloads. Scale a demo service on queue length instead, then compare response times and pod usage against CPU-based scaling.
5. Slow query advisor for PostgreSQL — Python, PostgreSQL, pg_stat_statements
   Small teams rarely have a database specialist. Parse EXPLAIN output from slow queries, suggest candidate indexes, test them on a copy of the data, and report the actual speed improvement.

Web and mobile application topics

1. Rotating QR attendance with geofencing — Flutter or React Native, Laravel or Django, JWT
   Static QR codes get forwarded over WhatsApp. Issue short-lived signed codes, add a location and device check, and run it in real classes to measure proxy attempts and failed scans.
2. Offline-first record app for field health workers — Flutter, SQLite, REST API, background sync
   Patchy connectivity causes lost entries in rural work. Store records locally, sync when online, and design conflict handling for two workers editing the same household record.
3. Real-time collaborative code editor — React, Monaco Editor, Yjs, WebSockets
   Simultaneous editing breaks naive save logic. Use CRDTs so that edits merge without a central lock, then test convergence under simulated delay and disconnection.
4. Verifiable academic certificates using blockchain — Solidity, Hardhat, IPFS, React
   Fake certificates are slow to check by hand. Store only a document hash on a test network, keep personal data off-chain, and let employers verify a PDF in seconds.
5. Prerequisite-aware course and elective planner — Python, Neo4j or NetworkX, React
   Students discover clashes too late. Model subjects as a directed graph, detect impossible plans, and suggest elective sequences that meet credit rules and timetable limits.
6. Hospital queue simulator and appointment planner — Python, SimPy, Django, Chart.js
   Long OPD waits often come from poor slot design. Simulate arrivals, test different slot patterns, and turn the best result into a booking system with live wait estimates.

Applied AI and data topics

1. AST-based source code plagiarism detector — Python, tree-sitter, scikit-learn
   Renaming variables defeats simple text comparison. Compare program structure through abstract syntax trees, and evaluate against deliberately disguised copies of lab assignments.
2. Marathi and Hindi voice complaint portal — Whisper, Python, IndicBERT, FastAPI
   Typed forms exclude many citizens. Transcribe spoken complaints, classify them by department, and measure where recognition fails on real regional accents and background noise.
3. College notice assistant with cited answers — Python, embeddings, ChromaDB or FAISS, an LLM API
   Circulars pile up in PDFs nobody can search. Build a retrieval pipeline that answers from official documents with a source link, and test how often it refuses when the answer is missing.
4. Federated learning across simulated hospitals — Python, PyTorch, Flower
   Hospitals cannot easily pool patient data. Train across simulated sites on a public dataset, make the data distribution uneven, and compare results with central training.
5. Resume-to-job matching with explanations — Python, spaCy, sentence-transformers, Laravel
   Keyword filters reject good candidates and tell them nothing. Parse resumes, match skills against job descriptions using embeddings, and show which skills drove each score and which are missing.

Choosing between these as a team

IT students often lean towards deployed applications and cloud work, while computer engineering students may have more room for algorithms and systems internals. Neither is a rule, so check what your guide and department expect.

• Match the topic to skills someone on the team already has, and limit new technology to one major item.
• Check that the data or hardware is available this week, not just in theory.
• Find two or three recent papers on the problem so your literature survey has a base.
• Divide the work into modules that each member can explain alone.
• Decide what you will compare against, because a baseline makes your results meaningful.

If you want guidance along the way

Road2Tech, based in Pune, mentors students through topic selection, reference papers, synopsis, code and documentation, with the aim that you understand and can explain what you submit. Online support is available through AnyDesk or TeamViewer.