• Developed predictive models using Random Forest, GBM, and LSTM in TensorFlow and Scikit-learn, increasing accuracy by 20%, and churn prediction models with XGBoost and CatBoost, achieving 94% accuracy.• Automated feature extraction for time-series data with Azure Machine Learning and Python scripts, and streamlined model training and deployment with Azure Data Factory and Azure DevOps, cutting cycle times by 30%.• Deployed real-time models on edge devices with Docker and Kubernetes, and fraud detection models with TensorFlow Lite, reducing fraud incidents by 35%.• Integrated Azure Monitor and Grafana for real-time monitoring and alerts via Azure Logic Apps, enhancing service reliability and improving network performance by 15%.• Processed real-time 5G network data using Apache Kafka and Azure Blob Storage, and architected data pipelines for datasets over 10 billion rows, ensuring high performance.• Reduced network downtime by 20% and improved efficiency by 20% with optimized ML models and Python/PyTorch solutions.• Enhanced anomaly detection by 20% with a TensorFlow-based autoencoder, and reduced model maintenance time by 25% with automated Azure ML pipelines.• Led and mentored a data science team, cut cloud infrastructure costs by 25% with Azure solutions, and managed ML lifecycle with MLflow and TensorFlow Serving for consistent performance.

• Built and deployed predictive models using XGBoost and RNNs to forecast player churn and spending, enhancing player retention by 20% and increasing in-game purchases by 15%. Developed recommendation systems with Collaborative Filtering, driving personalized content and boosting engagement and retention.• Created AI models using Deep Q-Learning for dynamic difficulty adjustment, improving game responsiveness and increasing player satisfaction by 25%. Applied Isolation Forests, Autoencoders, and CNNs to detect abnormal gameplay, reducing cheating incidents by 40%.• Leveraged Apache Kafka and Flink for real-time processing of game telemetry data, improving anomaly detection and game performance. Automated data pipelines and model training with Apache Airflow and Docker, enhancing productivity by 25%• Conducted A/B tests using Google Optimize, leading to feature developments that increased player satisfaction by 10% and revenue by 15%.• Integrated ML models into the game engine and collaborated with developers to enhance AI capabilities. Managed ML model deployment and infrastructure with Docker, Kubernetes, and Azure Monitor, ensuring scalability and reliability.• Developed and maintained efficient game development pipelines with Azure Data Factory and GitHub Actions, and provided production support for ML projects to ensure minimal downtime and quick issue resolution.• Designed and deployed MLOps services using Docker Compose stacks, including MLFlow and workflow orchestration for game analytics. Led the creation of an on-prem MLOps platform for seamless ML model training, deployment, and monitoring.• Implemented NLP techniques to enhance text classification, increasing content accuracy and relevance by 20%. Deployed ML models for predictive analytics in game design, achieving a 15% improvement in player engagement and game balance.• Built and monitored ML models with Azure Pipelines, ensuring reproducibility and fault tolerance in game analytics.

• Processed large-scale transactional data using Apache Spark, Hadoop, and related tools (Sqoop, Pig, Flume, Hive, MapReduce, HDFS) for efficient handling of high-volume financial data.• Utilized SQL and Python for data analysis and customer segmentation, optimizing marketing strategies and improving operational efficiency by 15% and transaction processing speed by 10%.• Built real-time data pipelines with Apache Flink, Kafka Streams, and AWS services (Lambda, EC2) for enhanced transaction monitoring and fraud detection.• Developed fraud detection models using Random Forests, GBM, Isolation Forests, Autoencoders, and Logistic Regression, improving fraud detection accuracy by 25% and reducing fraud rates by 35%.• Managed credit data with Pandas and NumPy, developing credit risk models using Logistic Regression and SVMs, which reduced default rates by 10% and improved credit risk assessment accuracy by 22%.• Applied K-Means, Hierarchical Clustering, and Collaborative Filtering for customer segmentation and recommendation engines, optimizing marketing strategies and increasing ROI by 20% through refined A/B testing.• Engineered and deployed machine learning algorithms, transitioning models from research to production. Developed a multi-label classification neural network in PyTorch for linking loan claims and data.• Built real-time dashboards with Grafana and created interactive dashboards in Tableau and Power BI, improving transaction flow monitoring and system performance.• Utilized AWS Redshift for managing transaction data, integrated HBO consumer subscription data from AWS SQS into Snowflake and Postgres using AWS Boto3 API.• Developed a stability metric for model output drift using Jaccard index and Kernel density estimation, and created an Azure pipeline for it, improving model monitoring and stability.• Collaborated with data engineers, ETL teams, and clients to streamline data acquisition, schema design, and troubleshooting.

• Managed and processed clinical trial datasets using SQL and SAS, ensuring accuracy and efficiency in data management.• Built predictive models (e.g., Logistic Regression, Decision Trees) to forecast patient responses and identify adverse events. Developed additional models (e.g., Random Forests, SVMs) to predict drug efficacy and side effects, enhancing drug development decisions.• Applied Kaplan-Meier and Cox Proportional Hazards models to analyze survival rates and treatment efficacy, improving trial efficiency by 15% and reducing patient dropout rates.• Integrated and cleaned data from various sources using Python and Pandas for comprehensive analysis. Developed automated data pipelines with Python and Apache Airflow, reducing data processing times by 30%.• Validated models using cross-validation, ROC AUC, and Precision-Recall Curves to ensure accuracy and reliability.• Improved drug candidate selection by 20%, reducing time and cost in the drug development pipeline through enhanced data handling and predictive modeling.• Conducted analysis with R and Python to identify patient subgroups for personalized medicine. Used K-Means and Hierarchical Clustering to group patients based on genetic markers and clinical profiles.• Collaborated with biostatisticians to tailor treatment plans, improving outcomes by 15%, and supported personalized medicine efforts with data-driven segmentation and treatment recommendations.• Combined data from EHRs, insurance claims, and patient registries for robust post-market surveillance and comprehensive analysis.• Created visual reports and dashboards in Tableau for non-technical stakeholders, facilitating better understanding and decision-making.• Worked with cross-functional teams to integrate data science insights into drug development and patient care, enhancing overall project outcomes.

• Led the risk analytics team in developing and implementing models to assess credit risk, using advanced statistical techniques to reduce loan default rates by 20%.• Optimized loan approval processes by applying statistical methods and predictive analytics, resulting in an 18% increase in loan approval rates.• Developed a credit risk scoring model using Python and Logistic Regression, achieving a Kolmogorov-Smirnov statistic of 55% and capturing 7 times more defaults in the top decile compared to the bottom.• Published backend programming logic in SQL for internal credit ranking reports covering 25 million customers, creating a new revenue stream with a potential of $20 million.• Led a team of 4 developers to build a risk scorecard in Python for unsecured personal loan products, resulting in a 10% increase in approval rates at the same risk tolerance for an applicant base of 4.8 million people.• Implemented rule-based segmentation and targeting in Python in collaboration with the marketing team, achieving a 20% reduction in communication costs while maintaining the same click-through rate.• Designed an item-item collaborative filtering-based recommendation engine in Python for consumer durable products, achieving 65% prediction accuracy for the top 10 products customers are likely to buy in the next 3 months.• Constructed ranking logic for smartphones on the listings page using Python, based on price competitiveness and customer attributes, resulting in a 15% increase in smartphone revenue.• Designed a Python program to automate data extraction and preprocessing for a self-learning model based on user inputs, reducing manual touchpoints by over 50% and processing time from 4 days to 8 hours.• Developed and implemented risk models that ensured accurate and efficient credit risk assessment and decision-making, enhancing overall operational efficiency and effectiveness.