Smart Data Warehouse Limited · Canada (Remote)

• Built credit risk model for fintech client processing 2,000–3,000 weekly loan applications; trained XGBoost classifier in Databricks achieving 0.76 precision identifying loan defaults.
• Deployed churn prediction model for B2B SaaS client with 12K users; API processed 2–3M events monthly on AWS Lambda and helped reduce churn from 8% to 5.5% over six months.
• Built demand forecasting system for e-commerce client with 1,200 products across 3 warehouses; weekly forecasts reduced stockouts by 15% and overstock by 12%.
• Set up model monitoring tracking prediction drift; caught precision drop from 0.72 to 0.58 before sales team noticed, enabling rapid retraining with recent data.
• Automated monthly reporting for 8–10 clients using Python, reducing generation time from 3 days to 4 hours.
• Managed end-to-end deployment across AWS (S3, Lambda, EC2, RDS), Snowflake, Databricks, and FastAPI.

Nigerian Communications Commission · Nigeria

• Built data pipelines processing 150–200M daily call records from four major telecom operators, maintaining a 50TB+ database covering 18 months of submissions.
• Cut monthly compliance report generation from 5 days to 6 hours by automating data collection and validation, catching 12–15% of submissions with quality issues.
• Optimised PostgreSQL queries via indexing and table partitioning, reducing execution time from 45+ minutes to under 3 minutes for annual compliance reports.
• Built consumer complaint dashboard used by 15+ regulatory officers to process billing disputes under strict 48-hour response times.
• Recovered 40% of database storage during critical shortages by establishing automated archival and compression for historical data.