Dennis Tile

WHAT I DO
• Machine learning & model evaluation: classification and regression with scikit-learn, statsmodels and TensorFlow/Keras (LSTM, SVM, logistic regression). I benchmark baselines, measure precision/recall and ROC/PR curves, run error analysis, and design recall-first evaluations, including two-stage pipelines with human review.
• Data & geospatial analysis: Python (pandas, NumPy), SQL, EDA, feature engineering, GeoPandas, DBSCAN, folium and Google Earth Engine, delivered as clear visuals, interactive maps and concise policy briefs.
• Full-stack & real-time development: React 18, Vite, Three.js, Node.js, Express, Socket.IO and PostgreSQL, including schema design, row-level locking, atomic updates and audit trails.
• Security & payments: JWT rotation, bcrypt, TOTP 2FA, rate limiting, fraud-risk scoring, idempotent webhooks and multi-provider payment integration with retry/backoff.
• Delivery: Git, GitHub Actions CI, Playwright, Vercel, Supabase, Docker (basic) and Jupyter.
SELECTED WORK
• Analyzed 76,562 GPS-tagged tree-planting records across all 47 Kenyan counties; built an equity metric (trees per 10,000 people) that identified 10 high-impact counties and delivered maps plus a policy brief.
• Led model benchmarking on a text-classification project, comparing TF-IDF baselines with an LSTM using precision/recall curves and error analysis.
• Built a real-time multiplayer platform with a unified wallet, six payment providers and an anomaly-detection watchdog.
I communicate findings clearly, work independently across time zones, and I'm open to remote contract work in data science, AI expert roles, model evaluation and analytics.