Olufemi Y.

I build data and AI applications end to end, from the statistical model to the deployed web app. Recent work: FDA 510(k) device-study statistics, a multi-tenant health-analytics SaaS, and a citation-grounded RAG assistant for clinical teams. Most data projects need two people: one who gets the numbers right and one who ships the software. I'm trained as a statistician (M.Sc. Statistics, Biometry) and I build production apps in Python, Next.js and Supabase. The model and the product get designed together and validated properly. RECENT WORK • FDA 510(k) statistics for a wearable sleep-monitoring device: agreement analyses (Bland–Altman, Passing–Bablok) and Clinical Study Report sections for adult and paediatric studies. • Technical lead on a multi-tenant health-economics analytics web app: Next.js, Supabase with row-level security, data pipelines from national health sources, automated tests. • A citation-grounded clinical AI assistant: LLM retrieval over a knowledge graph of clinical guidelines, FastAPI backend, offline desktop build. • A calibrated clinical risk-scoring model, from raw hospital data to a calculator with PDF reports (survival models, penalised regression, isotonic calibration, Streamlit). • Multi-site energy and water forecasting with anomaly handling. The client extended the scope from one site to four. • Quantitative research on 333 municipal closed-end funds: stationarity testing, mean-reversion half-lives and z-score indexes in R. WHAT I BUILD • AI and LLM/RAG applications grounded in your own documents and data • Machine learning, predictive models and time-series forecasting • Data-heavy web apps and MVPs (Next.js, React, Supabase/PostgreSQL) • Biostatistics for clinical and research work: study design, power, agreement studies, survival analysis, reports • Dashboards and R Shiny applications STACK Python (scikit-learn, FastAPI, Streamlit), R and R Shiny, TypeScript, React, Next.js, Supabase/PostgreSQL, ChromaDB, LLM APIs, Vercel, GitHub Actions. HOW I WORK Every project starts with a short scoping milestone covering a data check, the approach and acceptance criteria. Delivery follows in milestones, with reproducible code and documentation. If the data can't support the question, I'll tell you early. Send the dataset or spec and the decision it should support. I'll reply with an approach and a milestone plan.