I'm a data analyst from Niger. I like digging into messy datasets to understand what's actually going on, and building small tools — some of them AI-based — around problems I see at home.

I'm currently open to new opportunities — feel free to reach out.

Everything below is a real, finished (or actively maintained) project, grouped by the month I started it — so you can see how my work has evolved over time. 🔒 marks a private repository (the link won't open for other visitors, but I'm happy to walk you through it).

• 🚨 Vigie — Mobile Money Fraud Watch — A fraud-detection engine for a mobile money wallet in Senegal. Before each payment goes through, it checks 17 simple rules and a machine-learning model, then decides: accept, send for review, or block — always with reasons a person can read ("new phone on this account, PIN reset 6 minutes ago, empties 99% of the balance"). In testing, all 28 fraud cases set off at least one alert. Comes with a dashboard and an API. The operator, Doxal Money, and all its data are made up.

• 🧹 Housekeeper Interview Helper — A phone app that helps families in Niamey interview a housekeeper or kitchen helper. Every candidate gets the same five questions, plus real-life situations drawn at random (oil catching fire, a banknote found under the bed…), and at the end you get a score and a list of what still needs checking before hiring. It's a decision aid, not a lie detector: the score never looks at nationality, religion or family situation, and nothing leaves the phone.
• ✈️ Plane Crash Survival Analysis — A deep dive into real records of 5,244 plane crashes going back to 1908: not just building a model, but stress-testing it. Given that a crash already happened, can you predict beforehand whether anyone survives? I built in checks to stop the model from "cheating" (accidentally seeing information that's only known after the fact), then reported the honest results even where they're humbling — the simple model beats the fancy one, and neither one understands the deadliest crashes at all. Comes with two write-ups: one in plain language, one for anyone who wants to audit the method line by line.

• 🏦 UEMOA Credit Bureau Simulation — A simulation of how banks in West Africa are supposed to share loan data, so no one accidentally lends money to someone who's already deep in debt elsewhere. I built the full chain: cleaning messy bank files, matching up records that belong to the same person even when their name is spelled differently every time, and turning the result into a credit score — then measured how accurate each step really is (99.96% of planted errors caught, 100% correct matches). Every bit of data is synthetic, generated by the program itself.
• 👥 RSU Household Deduplication — Django — A smaller, cleaner rebuild of the household-matching tool from Kobo RSU Niger below, this time as a web app with a review screen: two records never get merged automatically, a person always checks and confirms first. Built for a job interview, so every line is deliberately simple enough to explain out loud.
• 🏘️ Kobo RSU Niger — A full household data-collection pipeline for a Unified Social Registry in Niger, built on KoBoToolbox: a survey covering 124 questions across 8 sections, daily data-quality checks, and a name-matching tool tuned for how Sahelian names get spelled differently by different people. Ships with a built-in demo dataset, so the matching logic runs end-to-end with no KoBo account needed.

• 📊 BRVM Stock Tracker — A robot that logs, every day, the closing prices of the BRVM — the stock exchange shared by eight West African countries — archived back to 2015. I then tested a bunch of "smart" trading ideas against eleven and a half years of real data. The finding: none of them beat pure luck. What actually pays off, consistently, is dividends. Free web app, and it says clearly it's not investment advice.
• 📇 SIATeG Digital Business Card — A digital business card built for a Niger business forum: show a QR code instead of handing out paper, and the other person's contact lands straight in their phone in two seconds. Works offline, nothing is ever sent to a server.
• 🚗 DiagAuto Niger — A small app to help you understand a car problem before paying a mechanic to tell you. Type in the error code your scanner shows (or just upload a photo of the screen), and AI explains what it likely means, how serious it is, and what to check first, cheapest checks first.
• 📈 Statys — A multi-tenant platform that lets an analyst explore a dataset without writing any code. You drop in a CSV or Excel file, and it walks through each variable — univariate and bivariate analyses, with the right statistical test picked for you — then exports the whole thing as a PDF report. Node.js on the front, FastAPI on the back.
• 💬 Niger Fintech Reviews — I scraped the Google Play reviews of Niger's mobile money and banking apps to see what people actually complain about. After cleaning them up, I tagged the recurring themes, ran sentiment on them, and put it all in a dashboard with a short write-up of what it means for the apps.
• 🗳️ Niger 2020 Election Analysis — A close look at the first round of Niger's December 2020 presidential election, commune by commune — all 266 of them. I turned the raw CENI results into clean datasets, explored them in a notebook, and built a single-file HTML dashboard where you can read the winner and turnout off a map, filter by region, and search the table.
• 🌊 Niger Flood Early Warning — A small early-warning system for floods in Niger. It scores each department's risk from open climate data (rainfall, vegetation via satellite) and the flood impacts humanitarian agencies have recorded over the years.
• 💳 Mobile Money Fraud Detection — Catching fraudulent mobile money transactions in a simulated transactions dataset, from cleaning the data in Python through the SQL queries to a dashboard built with Tableau and Chart.js.

• 🍽️ Niamey Restaurants Analysis — If you wanted to open a restaurant in Niamey, where would you put it, and what kind? I scraped ~490 places off Google Maps and dug in: cleaning with Python and DuckDB, a few statistical tests, geospatial clustering on maps, and a scoring model to rank neighborhoods against each other.
• 📡 MTN Churn Analysis — Why do MTN Nigeria customers leave? I looked at 974 records (496 customers) with Python, PostgreSQL, and Tableau. Churn lands at 29.2%, and the report points to the segments worth focusing retention on.
• 🛒 Customer Behavior Analysis — Digging through 3,900 shopping records to understand what drives customer behaviour, from exploring the data in Python and PostgreSQL to a Power BI dashboard and a written report on segments, revenue, discounts, and loyalty.

• 📄 🔒 ATS — AI Resume Screener (private) — CV screening without the tedious part: you paste a job description, the app reads a Google Drive folder of resumes and ranks candidates by how well they fit, using Gemini. No database, nothing stored — just TypeScript.

• 🪑 Market Scanner Niger — Gaskiyar Kaya 🇳🇪 ("the truth about the goods" in Hausa) — your pocket carpenter for second-hand furniture in Niamey. Snap up to 5 photos of a sofa or wardrobe at the market or on Facebook, type the seller's price, and the AI tells you what it's really made of, what it's worth in FCFA, whether it's a good deal, and how far to haggle. You can send the summary to a friend on WhatsApp. Try the app.

🌱 More projects on the way.

Focus areas: Data Analysis · Data Visualization · SQL & NoSQL Databases · Data Wrangling · Web Scraping · Geospatial Analysis · Statistical Testing · AI-assisted tools

• 💼 LinkedIn: linkedin.com/in/mohassane

_{💬 Open to opportunities in data analysis.}