Bhuvaneswar Sarakanam

Results-driven AI/ML Engineer with 4+ years designing and deploying production machine learning systems across insurance, manufacturing, and healthcare. Specialist in Generative AI (GenAI) building Retrieval-Augmented Generation (RAG) pipelines, Large Language Model (LLM) integrations, and agentic workflows using LangChain and LlamaIndex. Proven MLOps track record on AWS SageMaker, Kubeflow, and Kubernetes, consistently delivering 20–40% efficiency gains at 99.9%+ production uptime.

Experienced across the full ML lifecycle feature engineering, model training, drift detection, automated retraining, and multi-cloud deployment (AWS, Azure, GCP). MS in Computational Data Science, Purdue University.

AI/ML Engineer - New York Life Insurance - Remote, USA

(2024-10)

• Engineered fraud and anomaly detection models (XGBoost, Autoencoders) on insurance claims data, cutting false positives by 30% and reducing investigator turnaround time by 25% directly improving claims throughput.
• Deployed an LLM-powered document intelligence pipeline (AWS Textract + GPT-4) to auto-extract and classify unstructured claims documents, eliminating 35% of manual underwriting review time and accelerating policy decisions.
• Architected end-to-end MLOps infrastructure on AWS SageMaker, Triton Inference Server, and Kubernetes, achieving 99.95% model serving uptime and cutting retraining overhead by 50% through pipeline automation.
• Implemented proactive model drift detection and automated retraining workflows, increasing fraud recall by 20% over six months by continuously adapting to evolving fraud patterns.
• Built Airflow-orchestrated ETL pipelines ingesting millions of daily insurance records at 99.9% data accuracy, enabling real-time underwriting decisions and downstream model freshness.
• Embedded Explainable AI (XAI) frameworks into production fraud models to support regulatory compliance and audit transparency, enabling stakeholders to directly interpret model predictions; mentored 2 junior ML engineers in best practices.

AI/ML Engineer - Tata Consultancy Services - Hyderabad, India

(2021-10 - 2023-06)

• Built predictive maintenance models (LSTM, XGBoost, Random Forest) on Internet of Things (IoT) time-series data (~10M records), reducing equipment downtime by 30% and improving maintenance scheduling by 25%.
• Designed rolling-statistics and lag-variable feature sets for time-series forecasting, lifting model accuracy by 15% to 92% overall while reducing false positives by 20% outperforming the baseline significantly.
• Designed Azure ML + Kubeflow + MLflow end-to-end pipelines for model training, validation, and deployment, cutting deployment time by 40% and enabling full experiment reproducibility.
• Applied HuggingFace Transformer-based Natural Language Processing (NLP) models to classify unstructured maintenance logs, improving fault categorization accuracy by 18% over rule-based approaches.
• Containerized and deployed ML models via Docker with ERP system integration and automated retraining pipelines, reducing manual intervention by 50% and enabling zero-downtime production rollouts.

ML Engineer - Omega Healthcare Management Services - Andhra Pradesh, India

(2020-01 - 2021-09)

• Built and deployed supervised classification models (Logistic Regression, Decision Trees, Random Forest) for patient readmission risk scoring and claims fraud detection, achieving up to 85% prediction accuracy across both use cases.
• Applied NLP techniques including TF-IDF and Bag of Words to classify clinical text and insurance claims data at 78% accuracy, automating previously manual documentation review workflows.
• Built FastAPI and Flask RESTful APIs for real-time model inference, reducing prediction latency by 20% and accelerating clinical decision-making.
• Maintained production model health on AWS SageMaker with MLflow-tracked automated retraining pipelines, sustaining 99% uptime and ensuring continuous performance across all clinical prediction services.
• Boosted average F1 Score by 10% across all predictive models through systematic hyperparameter tuning (Grid Search, Random Search) and rigorous cross-validation aligned to healthcare Key Performance Indicators (KPIs).

Master of Science - Computational Data Science - Purdue University (2023-08 - 2025-05)

Bachelor of Engineering - Electronics & Communication - Adikavi Nannaya University (2017-06 - 2021-08)