I'm Gbolahan Bantefa, a Software Engineering student at Babcock University. I'm passionate about machine learning, backend infrastructure, and full-stack development, utilizing technologies like Python, Java, TypeScript, AWS, and Supabase to build robust systems. I enjoy tackling complex data and automation challenges, from architecting explainable AI pipelines for credit-risk evaluation to building full-stack inventory platforms and fraud detection engines.
(click to expand)
Babcock University, Ilishan-Remo B.Sc. Software Engineering | Expected May 2028
- Coursework: Computing in Python, Math and Statistics, Web and Mobile Application.
Document Tamper Detector
Engineered a multi-layer digital document forensics engine that detects pixel manipulation, text splicing, and GenAI inpainting with 94%+ tamper localization accuracy using Error Level Analysis (ELA) and Laplacian noise grain filtering. Built an interactive forensic dashboard with side-by-side tamper heatmaps to accelerate fraud audit workflows to under 2 seconds.
Technologies: OpenCV | NumPy | PyZbar | Streamlit | Hugging Face Spaces
Payment-Failure Prediction & Smart Routing
Achieved 0.88-0.90 ROC-AUC predicting transaction failures by training an XGBoost model enriched with GraphSAGE corridor-reliability embeddings. Cut simulated failed-transaction retries by 30% via an LLM tool-calling agent that recommends alternate routes.
Technologies: XGBoost | GraphSAGE | FastAPI | ONNX Runtime | Docker | SHAP
Automated Inventory Control System
Led full-stack development of an inventory management system featuring authentication, real-time UI updates, notifications, and analytics dashboards. Designed a secure, scalable architecture using JWT-based authentication and JSONB data modeling.
Technologies: React | TypeScript | Supabase (PostgreSQL) | Hono | REST APIs
(click to expand)
Verakri (Self-Directed Venture) | May 2026 - Present
- Lifted ROC-AUC to 0.78 by training a LightGBM model on 300K credit records + engineered alternative-data features.
- Delivered auditable lending decisions via instant SHAP reason codes by building an explainable AI pipeline with an LLM assistant.
- Shipped a production-grade credit-risk product with sub-500ms prediction latency deployed via FastAPI + ONNX in Docker.
LightGBMFastAPIONNXDockerMachine Learning
INNOVACORE (Ogun, Nigeria) | Oct 2025 - Present
- Led the development of a full-stack Automated Inventory Control System with React, TypeScript, Supabase, and a serverless backend.
- Implemented inventory workflows including stock updates, low-stock alerts, and secure per-user data isolation.
ReactTypeScriptSupabaseServerless
(click to expand)
| Organization | Certification / Role | Details |
|---|---|---|
| NVIDIA | Fundamentals of Accelerated Data Science | Mastered techniques to scale data manipulation and ML algorithms natively on GPUs using RAPIDS. |
| Deloitte Australia | Technology Job Simulation on Forage | Completed development/coding simulation and wrote a dashboard creation proposal (Sept 2025). |
Building the future, one model at a time.