Samad L

- Engineered credit risk prediction models using PyTorch and XGBoost, analysing two million customer financial records, improving default detection accuracy by 26% and reducing loan review time by 35%. - Analysed ingestion and feature engineering pipelines for transaction datasets in Snowflake and Python, processing over ten million daily records, improving data readiness for machine learning workflows by 40%. - Arranged anomaly detection algorithms for fraud monitoring using Isolation Forest and LightGBM, identifying 18% more suspicious transactions across five million monthly card payments, strengthening operational security controls. - Implemented explainable AI (XAI) solutions for lending decisions using SHAP and LIME, improving model interpretability and enabling regulatory compliance reviews across three mission-critical credit products. - Created interactive dashboards in Looker and Power BI to visualize customer risk profiles, supporting 25+ analysts in strategic decision-making and reducing report generation time by 50%. - Defined with DevOps teams to containerize and deploy ML models using Kubernetes and Docker, enabling scalable real-time recommendation engines, and decreasing application latency by 28% across digital banking platforms.

- Prepared and implemented deep learning models using TensorFlow and Keras to predict patient readmission, improving prediction accuracy by 28% across four healthcare provider datasets with 1.2 million records. - Revamped automated ETL pipelines in Python and Apache Airflow for integrating 10+ heterogeneous clinical and operational datasets, reducing data processing time by 45% and ensuring HIPAA-compliant data handling. - Authored with data scientists to deploy NLP models on clinical notes, extracting over 200K medical entities to enhance disease risk profiling and improve early diagnosis reporting by 22%. - Built interactive healthcare dashboards in Power BI, visualizing AI model outputs and key clinical KPIs, enabling hospital administrators to make decisions 35% faster across three major hospitals. - Optimized convolutional neural networks for medical imaging classification, reducing inference time by 30% while maintaining 94% accuracy across 50K+ radiology images for diagnostic support systems. - Deployed AI models on Azure ML using containerized microservices with Docker, supporting real-time patient monitoring applications and reducing model deployment cycle time by 40% for hospital integration.

- Developed and deployed machine learning models using Python and Scikit-learn, improving client sales forecasting accuracy by 22% across 5+ retail projects, reducing overstocking and stockout incidents. - Proposed data preprocessing pipelines using Pandas and NumPy, processing over one million transactional records weekly, enhancing data quality by 35% and reducing manual cleaning effort by 50%. - Collaborated with cross-functional teams to implement predictive analytics solutions on customer churn, achieving a 15% reduction in churn rates for telecom clients by leveraging classification algorithms. - Built NLP-based sentiment analysis models using NLTK and spaCy, analysing 500K+ customer reviews to generate actionable insights, increasing marketing campaign efficiency by 18%. - Conducted feature engineering and hyperparameter tuning for regression and classification models, improving model performance metrics (accuracy and RMSE) by an average of 20% across seven client datasets. - Designed and maintained Liaised ML model dashboards in Tableau, visualizing KPIs and prediction trends for stakeholders, improving decision- making speed by 40% and adoption across three client teams.