Data final year project ideas
Data final year project ideas from the free Sefism library, each scoped for a single semester with a suggested stack and deliverables.
- 300 project ideas
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
70 ideas · page 1 of 3
Next 48-Hour PM2.5 Forecasts for South Asian Cities from Open Sensors and Weather Data
Each winter, smog in Lahore, Delhi, Dhaka and Kathmandu closes schools and harms health, and decisions are made with little warning. Build a model that forecasts PM2.5 for the next 48 hours at city sensor locations, using open sensor readings, weather forecasts and fire detections, and compare it against the global CAMS forecast. Present it as a simple outlook a school or clinic can use to plan.
Which LLM Understands Roman Urdu and Bangla Business Messages? A Three-Model Evaluation Harness
Small businesses in Pakistan and Bangladesh get orders and complaints as WhatsApp-style messages in Roman Urdu, Urdu script, Bangla and English mixed together. Build an evaluation harness that compares three LLMs (for example one hosted frontier model, one cheaper hosted model and one open model) on real-shaped tasks such as extracting order details and classifying complaints, with a human-checked gold set, cost per correct answer and a public results page.
Model Card and Datasheet Generator from Training Runs
Model cards and dataset datasheets are the standard way to document what a model does, what data it learned from and where it fails, but they are written by hand at the end, if at all. Build a generator that pulls facts from MLflow runs, dataset files and evaluation results, drafts a model card and datasheet, flags missing or unsupported sections, and keeps the card in sync as the model is retrained.
Pipeline That Turns Provincial Statistical PDFs into a Versioned Open Dataset
Governments publish important statistics as PDF yearbooks and reports: schools, hospitals, crops, prices and population by district. Build a data pipeline that extracts tables from these PDFs, cleans and checks them, tracks every change between editions, and publishes them as tidy, documented, versioned datasets that journalists, students and researchers can query directly.
District Medicine Stockout Early Warning from SMS Stock Reports at Rural Clinics
Rural clinics run out of essential medicines while a clinic 20 kilometres away has boxes about to expire. Build a system where clinic staff report stock on hand by structured SMS or a tiny web form, and a district dashboard forecasts which clinics will stock out in the next weeks and suggests transfers between clinics. The focus is the district view and forecasting from patchy reports, not a single pharmacy's ledger.
Mobile Money Agent Cash and E-Float Forecasting to Prevent Turned-Away Customers
Mobile money agents in Kenya, Pakistan, Bangladesh and Nigeria lose customers when they run out of cash or electronic float, especially on salary days and before holidays. Build forecasting models for each agent's cash and e-float needs from simulated transaction histories with realistic calendars, and a simple app that tells the agent and the aggregator when and how much to rebalance. Measure how many turned-away transactions the forecasts would prevent.
Glacial Lake Growth Monitoring from Sentinel-2 Imagery for Outburst Flood Risk in Northern Pakistan
Glacial lakes in Gilgit-Baltistan and Chitral are growing as glaciers melt, and some can burst and flood villages downstream. Build a pipeline that maps lake outlines from free Sentinel-2 satellite images every month, tracks area change over several years, flags lakes growing fast, and lists the settlements downstream. Validate the outlines against published inventories and hand digitising.
FOCUS Format Converter for Smaller Cloud Providers and AI API Bills
The FinOps Open Cost and Usage Specification (FOCUS) gives cloud bills a common format, and the big clouds now export it. Startups also pay smaller hosts, AI API providers and SaaS tools whose bills arrive in their own formats. Build converters that turn those bills into valid FOCUS data, a validator, and a dashboard that shows total cost and unit cost (per customer or per feature) across every provider in one place.
Tool Selection Evals and Description Linter for MCP Servers
An MCP server can work perfectly and still fail in practice because models pick the wrong tool or fill its arguments badly. Build a tool that lints MCP tool names, descriptions and schemas against good-practice rules, generates realistic user requests for each tool, measures how often several models choose the right tool with valid arguments, and suggests rewrites that improve the score.
Mandi Arrival and Price Forecasting with Sell-or-Store Advice for Farmers
Farmers often sell right after harvest when wholesale market (mandi) prices are lowest. Build forecasts of arrivals and prices for a few crops at nearby markets from public government data, and turn them into simple advice: sell now, or store for a few weeks, after storage costs and losses. Test the advice by replaying past seasons and measuring how often it would have helped.
Inter-Annotator Agreement and Adjudication Tool for Data Labelling Teams
AI depends on labelled data, and labelling teams in Pakistan, India, Kenya and elsewhere produce a lot of it. Build a quality layer for labelling projects that measures agreement between annotators, spots unreliable labels and confusing guidelines, routes disagreements to an adjudicator, and estimates true labels from noisy votes, working with exports from Label Studio and similar tools.
Cyclone Shelter Capacity and Evacuation Route Planner for Coastal Bangladesh
When a cyclone warning is raised, coastal families must reach a shelter in time, but some shelters overflow while others sit half empty, and some villages are too far from any shelter. Build a planning tool that uses open maps and population data to assign villages to shelters within walking time and capacity, finds coverage gaps, and suggests where new shelters would help most. Test it on one coastal upazila.
Drone Plant Stand Counting and Gap Mapping for Smallholder Maize and Sugarcane Fields
A few weeks after sowing, farmers need to know how many plants came up and where the gaps are, so they can resow in time. Build a pipeline that takes low-altitude RGB drone photos of a small field, stitches or tiles them, counts plants with a detection or density model, and produces a gap map and an estimate of plant population per acre. Compare it against hand counts in marked sample plots.
Remittance Cost Comparator for Gulf to South Asia Corridors Using World Bank Price Data
Workers in Saudi Arabia, the UAE and Qatar send billions home to Pakistan, India, Bangladesh and Nepal, and the true cost hides in the exchange rate margin, not the fee. Build a comparator that uses the World Bank's Remittance Prices Worldwide data plus user-entered quotes to show the real amount received, explains fee versus margin, and tracks corridors against the international 3 percent cost target.
River Flood Forecasting with LSTM Models from Upstream Gauges and Satellite Rainfall
Train and evaluate machine learning models that forecast river discharge one to five days ahead for a few South Asian or African river basins, using upstream gauge readings, satellite rainfall and public hydrology datasets. Compare against simple persistence and published global forecasts, and report the lead time at which a warning threshold would have been crossed in past floods. This is the forecasting model, not a sensor network or an alert app.
Urdu and Bangla Embedding Models and Vector Databases: A Retrieval Benchmark
Teams building search or RAG in Urdu and Bangla pick an embedding model and a vector database by guesswork. Build a reproducible benchmark that measures retrieval quality for several multilingual embedding models on Urdu, Roman Urdu and Bangla queries, and then measures recall, latency, memory and filtered search across several vector stores with the same data, so a developer can choose with evidence.
One App, Three Databases: Benchmarking PostgreSQL, MongoDB and SQLite on a Real School Workload
Students pick a database because a tutorial used it. Build the same school management backend (admissions, attendance, fees, results) on PostgreSQL, MongoDB and SQLite, generate a realistic workload, and compare correctness under concurrency, query speed, schema change effort, backups and hosting cost. The result is an evidence-based guide to choosing a database for a typical local business app.
Data Contract and Schema Drift Checker for Event Pipelines
When an app team renames a field or changes a type, the analytics pipeline downstream breaks quietly and dashboards show wrong numbers for weeks. Build a tool that lets producers and consumers agree a data contract in a standard YAML format, checks every pull request and every batch of events against it, explains breaking changes before they ship, and tells the right people what will break.
Mule Account Detection with Graph Machine Learning for Mobile Wallets
Build a simulator of a mobile wallet payment network with realistic scam and mule account patterns, then train and compare graph machine learning models that flag accounts used to collect and move stolen money. Scam proceeds in Pakistan often pass through chains of wallet accounts opened with borrowed or bought identities. Rules catch known patterns; the project tests whether learning from the shape of the money flow catches more, and explains each flag to an analyst.
Wheat Crop Monitoring and District Yield Estimates from Sentinel-2 for Punjab
Build a pipeline that maps wheat fields across Punjab from free Sentinel-2 satellite images, tracks how the crop develops through the season and estimates district yields weeks before harvest. Wheat is Pakistan's staple, and procurement, import and flour price decisions depend on early estimates. The project tests how far open satellite data can go, with every estimate checked against published district statistics.
Open Data Portal Quality Scorecard with Freshness and Broken Link Checks
Build a tool that crawls government and humanitarian open data portals every week, checks whether each dataset can actually be downloaded, opened and used, and publishes a scorecard of freshness, broken links, file formats and missing metadata. Many South Asian portals list thousands of datasets, and a large share turn out to be stale PDFs or dead links. The scorecard tells portal teams what to fix and tells users what to trust.
MCQ Exam Item Analysis and Faulty Question Finder for University Teachers
Build a web tool where a teacher uploads the answer sheet results of a multiple-choice exam and gets back a report on every question: how hard it was, whether it separated strong and weak students, which wrong options nobody chose, and which questions probably have the wrong answer key. Many university and entry-test MCQs in Pakistan are marked on paper or OMR with results in Excel, so this analysis rarely happens. Fixing a few bad questions makes results fairer for every student.
Cloud Data Transfer Cost Attribution from Flow Logs for Small Teams
Build a tool that reads VPC flow logs and the cost and usage report from an AWS account, works out which service talks to which and over what path (same zone, cross-zone, NAT gateway, internet, cross-region), and attributes data transfer charges to the services and teams that cause them. Small teams see a large data transfer line on the bill and cannot tell where it comes from. The tool turns that line into a ranked list of flows with fixes.
Synthetic Patient Records Generator with a Privacy Attack Report for Hospital Data Sharing
Build a tool that takes a hospital table, generates synthetic records with several methods, and then attacks its own output to measure how much real patient information leaks. Hospitals want to share data with students and researchers, and synthetic data is often sold as automatically private. This project produces a report that shows both the usefulness and the privacy risk of each synthetic dataset before anyone shares it.
Browse project ideas by domain and difficulty
Each of these is its own page, so you can send someone straight to the AI ideas or the beginner ones.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
- 1:1 Mentorship
- Fresh Graduate Mentorship
- Career Guidance
Brand deals, promos and partnerships: tauseeffayyaz@gmail.com
Follow along
New writing, resources and project ideas land here first.