I developed a complaint escalation model and delivered its end-to-end MLOps on Azure ML. Each scoring pipeline run generates risk scores and runs monitoring with Evidently AI, keeping prediction delivery and model monitoring in the same workflow.
PythonXGBoostAzure MLEvidently AI
Behind the model +
Owned feature engineering, model selection, and validation, including performance assessment across complaints of different ages. Delivered predictions into operational systems to support proactive case prioritisation.
Early production monitoring showed that the highest-risk 30% of complaints captured 60% of escalations. This is an early observed result, not a guarantee of future performance.
EARLY PRODUCTION MONITORING
60%
of escalations captured in the highest-risk 30% of complaints
HIGHER RISKLOWER RISK →
Risk distribution illustrated for visual context.
MLOPS
01Train
→
02Validate
→
03Deploy
→
04Monitor
↳ CONTINUOUS FEEDBACK ──────────────────── ↵
Production ML, end to end.
Taking models into production is a core part of how I work. I built and now own a shared Azure ML framework that makes deployment, scoring, monitoring, and governance a repeatable part of delivery.
Azure MLMLflowObservability
Inside the framework +
Migrated models from Azure Synapse to Azure ML. Automated data movement, orchestration, scoring, and prediction delivery. Embedded model versioning, lineage, drift monitoring, and approval gates to support auditable delivery.
Secured production approvals through Cyber, Observability, and AI committee reviews, with ongoing ownership of design reviews and deployment approvals.
ANALYTICS
30hrs
returned to the team every month
More insight. Less manual work.
Python automation that freed up 30 hours each month, alongside predictive models and Power BI dashboards for better business decisions.
PythonPower BIPredictive analytics
Explore the approach +
Automated insights delivery with Python and developed proof-of-concept classification and regression models for budget planning. Built dashboards for KPI tracking and deep dives, and shared practical ML and statistics use cases with business teams.
THE PERSON BEHIND THE MODELS
An engineer’s mind. A scientist’s curiosity.
USAMA AZHAR DATA SCIENTIST · UK
I’m Usama, an electrical engineer turned data scientist. I care about what happens after a model leaves the notebook.
My work connects the full journey: understanding a business problem, finding the signal in the data, and building a system people can rely on. Today, I do that at ScottishPower in the UK.
An engineering background taught me to think in systems. Data science gave me a new way to improve them. I bring both to work that needs to be useful, measurable, and built to last.
