Home

Pavan Pratapagiri - Data Scientist / Machine Learning Engineer / Artificial Intelligence Engineer

[email protected]

Location: Richardson, Texas, USA

Relocation: Anywhere in usa

Visa: F1-STEM OPT

Resume file: Pavan Pratapagiri-Resume_1775570309213.pdf Please check the file(s) for viruses. Files are checked manually and then made available for download.

SUMMARY Data Scientist with 5+ years of experience building and operationalizing machine learning solutions for fraud prevention, risk assessment, and financial forecasting across insurance and capital markets. Specialized in Agentic AI frameworks, Retrieval - Augmented Generation (RAG), and deep learning using TensorFlow and PyTorch. Delivered measurable business impact, including 25% uplift in predictive accuracy and 40% reduction in fraud losses through cloud-based ML architectures on AWS and GCP (BigQuery, GCS). Strong expertise in distributed data processing (Spark, Hadoop), real-time decision systems, explainable AI (SHAP), and full ML lifecycle governance. TECHNICAL SKILLS Programming & Distributed Data Processing: Python (Pandas, NumPy, SciPy), Advanced SQL, PySpark, Apache Spark, Hadoop (HDFS, MapReduce), ETL Development, Data Modeling Machine Learning & Statistical Modeling: Scikit-learn, XGBoost, TensorFlow, PyTorch, ARIMA, LSTM, Time Series Forecasting, Anomaly Detection, Risk Modeling, Fraud Detection, Feature Engineering, Hyperparameter Tuning, Cross-Validation, RMSE, MAE, Precision-Recall, ROC-AUC Generative AI & Agentic Systems: Agentic AI Architectures, Retrieval-Augmented Generation (RAG), Large Language Models (LLMs), Transformer Models, Prompt Engineering, Embeddings, Vector Retrieval, Sentiment Analysis Cloud & Big Data Platforms: AWS (SageMaker, S3, EC2), Google Cloud Platform (GCP), BigQuery, Google Cloud Storage (GCS), Distributed Data Processing, Scalable Model Deployment MLOps & Model Governance: MLflow, Model Deployment, Model Monitoring, Drift Detection, SHAP (Explainable AI), CI/CD Pipelines, Docker, Kubernetes, REST APIs Streaming & Real-Time Systems: Kafka, Event-Driven Architectures, Real-Time Inference Pipelines EXPERIENCE Assurant, Texas, USA Jul 2025 Present Data Scientist Directed end-to-end development of XGBoost and TensorFlow models for vehicle claim risk scoring and fraud analytics, achieving 25% accuracy gains and decreasing false positives by 30% based on ROC-AUC and RMSE evaluation. Built event-driven risk assessment pipelines using PySpark, Kafka, and AWS SageMaker, shortening claim adjudication cycles by 35% across multi-state portfolios. Developed distributed ETL architectures with Spark and Hadoop to handle high-volume mobility and claims datasets spanning 50+ insurance programs. Advanced premium pricing methodologies through ARIMA and LSTM-based forecasting techniques, strengthening long-term revenue planning and actuarial insights. Implemented transformer-driven NLP and RAG frameworks to extract contextual intelligence from unstructured policy documents and customer communications. Integrated SHAP-based interpretability mechanisms and compliance controls to support audit readiness and transparent decision-making. Deployed containerized inference services with Docker and REST APIs while instituting automated drift detection and performance tracking workflows. Morgan Stanley, Texas, USA Jun 2024 Jun 2025 Data Scientist Constructed predictive frameworks for market risk evaluation and trading anomaly identification using XGBoost and TensorFlow, delivering 20% forecasting improvement and lowering unauthorized trading exposure by 40%. Designed streaming anomaly detection systems with PySpark and Kafka to supervise high-frequency transaction patterns and strengthen regulatory safeguards. Performed comparative modeling across ARIMA, LSTM, and gradient boosting techniques to anticipate forex and commodity price fluctuations for hedging optimization. Engineered high-throughput data processing workflows utilizing Spark and advanced SQL to accelerate quantitative research pipelines. Applied transformer-based sentiment modeling on earnings releases and financial news to refine short-term trading signals by 15%. Partnered with quantitative analysts and trading desks to integrate ML-driven insights into proprietary execution platforms, boosting operational throughput by 35%. University of Missouri, Kansas City, USA May 2023 May 2024 IS Technical Student Assistant Created and validated machine learning prototypes that elevated research model effectiveness by 30% across academic datasets. Produced SQL- and PostgreSQL-driven dashboards delivering analytical visibility to 20+ faculty and administrative stakeholders. Streamlined preprocessing pipelines with Spark and Pandas, decreasing preparation time for downstream modeling by 40%. Implemented forecasting and anomaly detection experiments within AWS SageMaker for research data validation. Developed Flask-based APIs enabling near real-time visualization and analytical reporting. Hexaware Technologies, India Aug 2020 Dec 2022 Machine Learning Engineer Built distributed ML workflows processing structured and unstructured healthcare data, increasing throughput by 40%. Developed predictive healthcare models using XGBoost and PyTorch, lowering diagnostic error rates by 25% and reaching 92% precision. Orchestrated feature engineering and SQL-based integration across heterogeneous clinical systems. Delivered secure inference services via Dockerized Flask applications deployed on AWS SageMaker, ensuring HIPAA-compliant data handling. Enhanced generalization performance through systematic validation strategies and hyperparameter optimization, cutting iteration cycles by 20%. EDUCATION Master of Science in Computer Science May 2024 University of Missouri, MO Bachelor of Science in Computer Science May 2021 Koneru Lakshmaiah University, India PROJECTS Topic Discovery & Trend Analysis on New York Times Articles (NLP, LDA) Technologies: Python, NLTK, SpaCy, Gensim, WordCloud, pyLDAvis Processed and normalized 50K+ news articles using advanced NLP preprocessing (lemmatization, stop-word filtering, token normalization) to improve topic coherence and semantic consistency. Implemented Latent Dirichlet Allocation (LDA) to identify 10 dominant editorial themes and evaluated model quality using topic coherence scores. Delivered interactive visualizations via pyLDAvis to enable exploratory trend analysis and interpretable topic insights. Real-Time Crowd Density Estimation via Semantic Segmentation Technologies: TensorFlow, YOLOv5 / Faster R-CNN, Flask, OpenCV Designed and deployed a dual-stage deep learning pipeline combining object detection and semantic segmentation for accurate crowd density estimation in high-traffic environments. Evaluated model performance using MAE, MSE, and F1-score, achieving high precision in dense scene detection scenarios. Operationalized inference through a Flask-based REST API to enable real-time video stream processing and scalable deployment. CERTIFICATION Microsoft Certified: Azure AI Engineer Associate Stanford (Coursera): Machine Learning Keywords: continuous integration continuous deployment artificial intelligence machine learning sthree rlang Missouri