Home

Stephen Gurram - Senior AI/ML Engineer

[email protected]

Location: Overland Park, Kansas, USA

Relocation:

Visa: Green card

Resume file: Stephen Gurram_1776095081169.pdf Please check the file(s) for viruses. Files are checked manually and then made available for download.

Client: KeyBank Pittsburgh, PA Role: Senior AI/ML Engineer Duration: May 2024 Present | USA Designed highly scalable machine learning and deep learning models using PyTorch (2.x) and TensorFlow (2.11+) for critical banking use cases including fraud detection, credit risk scoring, and transaction anomaly detection, improving detection accuracy by over 30% and strengthening enterprise risk mitigation. Architected and implemented end-to-end machine learning pipelines using Azure Machine Learning, Azure Data Lake Storage (ADLS Gen2), Azure Functions, and Azure Data Factory, enabling automated workflows for model training, validation, deployment, and orchestration, reducing deployment cycle time by approximately 40%. Built and deployed high-performance, real-time inference systems using FastAPI, Docker, and Azure Kubernetes Service (AKS), supporting millions of financial transactions with low latency, high throughput, and high availability. Developed advanced NLP solutions using BERT, RoBERTa, and Hugging Face Transformers (v4.x) for financial document classification, KYC/AML text analysis, and customer sentiment modeling, improving operational efficiency and decision-making. Designed and implemented Retrieval-Augmented Generation (RAG) pipelines using Azure OpenAI Service, LangChain, and Azure AI Search, enabling secure, context-aware responses over enterprise financial data and internal knowledge systems. Built scalable vector search solutions using Azure AI Search and FAISS, enabling semantic document retrieval for applications such as regulatory Q&A, policy search, and internal knowledge assistants. Developed LLM-powered intelligent assistants for banking use cases including customer support automation, document summarization, and compliance reporting, significantly reducing manual effort and improving response accuracy. Engineered end-to-end RAG workflows, including document ingestion, embedding generation, indexing, retrieval, and prompt orchestration, ensuring accurate and grounded LLM outputs. Implemented prompt engineering and response optimization techniques, reducing hallucinations and improving factual accuracy of LLM-generated outputs in regulated financial environments. Integrated security, governance, and access control mechanisms within AI systems, ensuring compliance with PCI-DSS, GDPR, and internal banking regulations. Implemented enterprise-grade MLOps pipelines using MLflow (2.x), Azure Machine Learning, Kubernetes (AKS), and Jenkins/Azure DevOps, enabling automated CI/CD, model versioning, monitoring, and lifecycle management. Engineered scalable data processing pipelines using Apache Spark (3.x) on Azure Databricks and Snowflake, optimizing feature engineering workflows and reducing data processing time by approximately 35%. Designed and deployed model monitoring and drift detection frameworks using Azure Monitor, Application Insights, Prometheus, and custom Python-based solutions to ensure model reliability and performance. Built and optimized recommendation systems using collaborative filtering and deep learning techniques, increasing customer engagement and product adoption rates by approximately 25%. Applied advanced statistical modeling, hypothesis testing, and A/B testing methodologies, ensuring measurable, data-driven business impact across banking use cases. Performed extensive hyperparameter optimization using Grid Search, Random Search, and Bayesian Optimization, improving model precision, recall, and overall predictive performance. Designed and implemented feature stores (Azure ML / Databricks Feature Store) and reusable ML components, improving model reusability, consistency, and accelerating development cycles. Collaborated with cross-functional teams including data engineering, risk analytics, product, and business stakeholders to deliver scalable AI/ML solutions aligned with enterprise goals. Integrated CI/CD pipelines using Git, Docker, Jenkins, and Azure DevOps, enabling faster, reliable, and auditable deployment of machine learning models. Supported and maintained mission-critical 24/7 production ML systems, ensuring high availability, rapid incident response, and minimal downtime in high-stakes financial environments. Client: CVS Health- Atlanta, GA Role: AI/ML Engineer Duration: Jan 2022 Apr 2024 | USA Developed predictive machine learning models using Scikit-learn (1.x) and TensorFlow (2.10+) for healthcare use cases such as patient risk stratification, hospital readmission prediction, and claims analytics, improving prediction accuracy by approximately 28% and enabling proactive clinical decision making. Designed and implemented scalable, production-grade data pipelines using Azure Data Factory, Azure Databricks, and Azure Data Lake Storage (ADLS), enabling efficient ingestion, validation, transformation, and orchestration of large-scale healthcare datasets for machine learning workflows. Built and optimized recommendation systems for personalized healthcare interventions and treatment suggestions using collaborative filtering and deep learning techniques, improving patient engagement and care outcomes. Engineered and deployed end-to-end machine learning systems using Docker, Azure Kubernetes Service (AKS), and Azure Machine Learning, ensuring high availability, fault tolerance, and scalability for real time healthcare applications. Performed extensive feature engineering and dataset construction using Pandas, NumPy, and SQL, including handling missing clinical data, normalization, encoding, and feature selection to improve model robustness and predictive performance. Developed and exposed RESTful APIs using Flask and FastAPI, deployed via Azure App Service and containerized environments, enabling low-latency, real-time inference for healthcare analytics systems. Utilized MLflow (2.x) integrated with Azure Machine Learning for experiment tracking, model versioning, and lifecycle management, ensuring reproducibility, auditability, and governance across ML systems. Leveraged Apache Spark (3.x) on Azure Databricks for distributed data processing and large-scale model training, reducing training time by approximately 35% and improving system efficiency. Applied advanced hyperparameter optimization techniques, including Grid Search, Random Search, and Bayesian Optimization, improving model generalization and performance across diverse healthcare datasets. Developed NLP-based solutions using spaCy (3.x) and Hugging Face Transformers, enabling automated clinical text classification, medical document summarization, and patient feedback analysis. Implemented robust model monitoring, logging, and alerting systems using Azure Monitor, Application Insights, and custom Python-based solutions, detecting performance degradation, data drift, and anomalies in production environments. Collaborated with cross-functional teams including data engineering, clinical analysts, and product stakeholders, translating healthcare requirements into scalable AI/ML solutions aligned with organizational goals. Built interactive dashboards and reporting tools using Power BI and Tableau, enabling visualization of model outputs, patient risk scores, and key healthcare KPIs. Ensured compliance with healthcare data regulations such as HIPAA, implementing secure data handling, access control, and audit-ready ML systems. Automated end-to-end machine learning workflows using Azure Machine Learning Pipelines, Azure DevOps, Jenkins, and Git, improving deployment efficiency, reliability, and reducing manual intervention. Client: GEICO Lakeland, FL Role: Machine Learning Engineer Duration: Sep 2019 Dec 2021 | USA Developed machine learning models for insurance claim fraud detection using Python (3.7/3.8) and Scikit learn, improving fraud detection accuracy by over 25% and reducing financial losses across high-volume claims processing systems. Designed and implemented scalable data processing pipelines using Apache Spark (2.x/early 3.x) and SQL, integrated with Azure Data Lake Storage (ADLS) for efficient storage and processing of large-scale insurance datasets, reducing data preparation time by approximately 30%. Built predictive models for claims severity estimation and policy risk scoring using TensorFlow (2.x) and Keras, supporting underwriting strategies and improving loss ratio performance. Developed NLP-based solutions using NLTK and early spaCy, enabling automated extraction of key information from claim reports, adjuster notes, and customer communications, improving operational efficiency. Engineered and deployed production-ready ML services using Flask-based REST APIs and Docker, hosted on Azure Virtual Machines and Azure App Service, enabling near real-time inference and integration with internal systems. Utilized Azure cloud services (Azure Blob Storage, Virtual Machines, and Azure Functions) for data storage, batch processing, and lightweight model deployment workflows in a scalable cloud environment. Performed comprehensive feature engineering and data preprocessing using Pandas, NumPy, and SQL, including handling missing values, encoding categorical variables, normalization, and outlier detection to improve model performance. Applied hyperparameter tuning techniques such as Grid Search and Random Search, improving model precision, recall, and F1-score for fraud detection and risk modeling tasks. Implemented basic model monitoring and logging mechanisms using Azure Monitor and custom logging solutions, tracking model performance and detecting early signs of data drift. Collaborated closely with data engineering teams to integrate ML workflows into existing ETL pipelines, leveraging Azure Data Factory for data ingestion and transformation. Built dashboards and reporting solutions using Power BI and Tableau, enabling stakeholders to monitor fraud trends, claims insights, and key performance metrics. Ensured compliance with insurance regulations and data privacy standards, maintaining secure handling of sensitive policyholder and claims data. Partnered with business analysts and underwriting teams to translate insurance requirements into scalable ML solutions aligned with business goals. Automated ML workflows including data ingestion, model training, validation, and deployment using Python scripting and Jenkins-based CI/CD pipelines, improving operational efficiency. Supported and maintained production ML systems in a 24/7 environment, ensuring high availability, rapid issue resolution, and minimal disruption to insurance operations. Client: Verizon Sandy, UT Role: Data Scientist / ML Engineer Duration: Jul 2017 Aug 2019 | USA Developed machine learning models using Python (3.6/3.7), Scikit-learn, and TensorFlow (1.x / early 2.0) for telecom use cases such as network anomaly detection, customer churn prediction, and service quality optimization, improving prediction accuracy and reducing customer attrition. Processed and analyzed large-scale telecom datasets (network logs, call detail records, customer usage data) using Apache Spark (2.x) on AWS EMR, enabling efficient distributed data processing and reducing data latency for analytics workflows. Built anomaly detection models using statistical methods and Scikit-learn, identifying network performance issues and service disruptions, leading to improved network reliability and reduced downtime. Designed and orchestrated data pipelines using Python, SQL, and Apache Airflow (1.x), ensuring reliable ingestion, transformation, and scheduling of high-volume telecom data for downstream analytics and ML applications. Developed time-series forecasting models using LSTM (TensorFlow/Keras) to predict network traffic patterns and demand fluctuations, improving capacity planning and resource allocation. Engineered and deployed machine learning services using Flask-based REST APIs and Docker, enabling scalable inference for internal telecom analytics platforms. Leveraged AWS services including EC2, S3, Lambda, and EMR for distributed data storage, batch processing, and model deployment, supporting scalable ML workflows in a cloud environment. Performed feature engineering and preprocessing using Pandas and NumPy, including handling missing data, normalization, and time-based feature extraction to improve model performance. Applied statistical analysis and hypothesis testing to validate models and support data-driven decision making for telecom operations and customer analytics. Built dashboards and visualization tools using Tableau and Power BI, enabling stakeholders to monitor network performance, customer trends, and key KPIs. Implemented model performance tracking and logging mechanisms, ensuring consistent accuracy and reliability of ML models in production. Collaborated with cross-functional teams including network engineers, operations teams, and business stakeholders, aligning ML solutions with telecom infrastructure and business objectives. Automated data workflows and model retraining pipelines using Python scripts and Jenkins-based CI/CD, improving efficiency and reducing manual effort. Ensured adherence to data governance, security, and telecom compliance standards, maintaining integrity and confidentiality of customer and network data. Supported and maintained production ML systems in a 24/7 environment, ensuring high availability, rapid issue resolution, and minimal disruption to critical telecom services. Client: Ruksun Software Technologies Ltd Pune Role: Junior Data Scientist / ML Engineer Duration: Jun 2016 May 2017 | India Developed machine learning models using Python (Scikit-learn) for client-focused analytics use cases such as customer segmentation and sales forecasting, improving prediction accuracy by 15 20% across multiple projects. Performed end-to-end data preprocessing, cleaning, and transformation using Pandas and NumPy, improving dataset quality and enabling more efficient model development. Designed and executed SQL-based data extraction and transformation workflows, preparing structured datasets for analytics and reporting across client engagements. Built and evaluated classification and regression models, applying appropriate algorithms and validating performance using metrics such as accuracy, precision, recall, and F1-score. Implemented foundational NLP solutions using NLTK and early spaCy, enabling text classification and keyword extraction from unstructured data sources. Created data visualizations and dashboards using Tableau and Matplotlib, delivering actionable insights to business stakeholders and supporting data-driven decision-making. Developed ETL workflows using Python, automating data ingestion from structured and semi-structured sources (CSV, JSON, relational databases). Collaborated with senior data scientists and cross-functional teams to improve model performance through feature engineering and basic hyperparameter tuning. Gained hands-on exposure to AWS services (EC2, S3) for data storage and experimentation, supporting scalable data processing and model execution. Assisted in deploying machine learning models via Flask-based APIs, gaining initial experience in integrating models into application workflows. Participated in Agile development processes, contributing to sprint planning, task execution, and timely delivery of analytics and ML solutions. Keywords: continuous integration continuous deployment artificial intelligence machine learning business intelligence sthree Florida Georgia Pennsylvania Utah