Balaga Raghuram

• Deploying a validated gradient boosting clinical AI model as a containerised REST API on Google Cloud Run (GCP), enabling scalable serverless risk prediction across NHS and privately commissioned hospital networks in the UK
• Managing the full MLOps lifecycle: model serialisation, Docker containerisation, version control via GitHub Actions, cloud infrastructure provisioning on GCP, and live production serving with automated monitoring
• Designing the REST API to accept structured patient clinical inputs and return real-time Low, Medium, and High deterioration risk scores for use by multidisciplinary oncology teams (MDT)
• Integrating Azure Application Insights for production monitoring and alerting, enabling real-time performance tracking and error detection post-deployment
• Ensuring full compliance with UK GDPR, the Caldicott Principles, and the NHS Data Security and Protection Toolkit throughout the production environment
• Skills: MLOps, GCP Cloud Run, Docker, REST API Design, Azure Application Insights, GitHub Actions, Clinical AI, Gradient Boosting, Python, NHS Data Governance

• Conducted advanced LLM evaluation and benchmarking using custom frameworks for large-scale experiments across multiple model architectures hosted on AWS and Azure ML
• Applied domain expertise to validate ML model outputs, ensuring technical accuracy and robustness of model behaviour across evaluation datasets
• Developed automated testing pipelines on Azure DevOps for algorithm assessment and validation under tight delivery timelines
• Skills: Hugging Face Transformers, LLMs, Python, NLP, ML Model Evaluation, AWS, Azure ML, Azure DevOps, Research Methodology

• Designed and built a clinical AI risk-stratification model across an NHS and private hospital network to identify cancer patients most at risk of deterioration following first-line treatment, across a cohort of 1,842 adult oncology patients
• Engineered 87 clinical features from electronic health records (EHRs), chemotherapy logs, pathology reports, radiology metadata, and lab results; managed 8.4% missing data using MICE imputation and deterministic NHS record linkage
• Benchmarked logistic regression, random forest, gradient boosting, and SVM using nested cross-validation and Bayesian hyperparameter optimisation with a prospective temporal train/validation split; gradient boosting selected with ROC-AUC 0.801, PR-AUC 0.609, C-index 0.791, and Brier Score 0.161
• Applied SHAP explainability (XAI) to surface the top clinical risk drivers: prior A&E attendance, nutritional decline, comorbidity burden, dose reductions, and advanced disease stage
• Validated findings using Kaplan-Meier survival analysis and decision curve analysis, demonstrating significant outcome separation across risk tiers (log-rank p < 0.001) over 36 months; delivered a Low/Medium/High risk stratification tool deployed across NHS and private hospital MDT pathways
• Maintained full compliance with UK GDPR, the Caldicott Principles, and the NHS Data Security and Protection Toolkit
• Skills: Gradient Boosting, SHAP, Survival Analysis, Feature Engineering, MICE Imputation, Bayesian Optimisation, Python, Scikit-learn, AWS S3 (data storage), GCP BigQuery (data pipeline), NHS Data Governance

• Led deployment of production-grade ML systems for 15+ enterprise clients, processing 2M+ property records annually at 99.2% data quality and a 95% deployment success rate
• Engineered 8+ ML models using TensorFlow, PyTorch, and ensemble methods, achieving 92% prediction accuracy and 18-25% uplift in client portfolio ROI
• Established end-to-end MLOps pipelines with CI/CD via GitHub Actions and Jenkins, automated monitoring, and drift detection sustaining 90%+ model performance in production
• Deployed models on Docker, Kubernetes, GCP Vertex AI, and Azure App Service, reducing deployment time by 40% and maintaining 99.9% uptime
• Integrated Azure Monitor and GCP Cloud Logging for real-time observability, enabling proactive detection of model drift and infrastructure anomalies
• Led a cross-functional team of 4 ML engineers from design to production, defining best practices for model development, testing, and code quality

• Delivered 25+ statistical and ML projects using regression, time-series analysis, and hypothesis testing across UAE and Middle East markets, processing datasets of 500K-1M+ records
• Built and deployed 12+ predictive models (TensorFlow, Keras, Scikit-Learn) achieving 92% accuracy and 18-25% improvement in client investment returns
• Implemented real-time model monitoring on GCP Vertex AI with automated drift detection, maintaining 90%+ performance over an 18-month production lifecycle
• Engineered end-to-end feature processing pipelines on AWS SageMaker and GCP Dataflow, reducing preprocessing time by 60% and improving F1-score from 0.78 to 0.91
• Created 30+ Power BI dashboards connected to Azure Synapse Analytics with 20+ KPIs, directly enabling 2M+ AED in client cost optimisation and revenue growth

• Co-developed the Statistics and Data Analysis course at IIT Tirupati, blending theoretical foundations with applied modelling and predictive analytics; designed and supervised hands-on projects applying ML algorithms to industry-style datasets
• At IIM Lucknow, performed SWOT analysis and statistical modelling for a DIPP government project using regression and time-series methods

• Recipient: UOB Merit Scholarship Award (2024), Department of Mathematics
• Distinction and scored 95 percent in computational statistics

Scored distinction

CGPA : 8.31 (83.1 Percentage)