Prasanna Pandey

Led end-to-end development and recalibration of over 20 ML models, improving predictive accuracy by up to 15% and scaling delivery across 40+ client environments via robust versioning and reusable pipeline design. Architected and implemented LLM-driven schema automation pipelines, reducing manual mapping by 60%, accelerating onboarding, and enhancing data consistency. Delivered a high-impact diagnostic prediction model identifying 80% of undiagnosed diabetes cases a year in advance, gaining national recognition and supporting early clinical interventions. Oversaw migration to Spark 3.3 on AWS EMR including comprehensive compatibility testing and refactoring to ensure zero downtime deployment and 25% reduction in training latency. Integrated ChatGPT into enterprise analytics workflows enabling privacy-aware prompt engineering and de-identification workflows. Mentored junior engineers, institutionalized ML knowledge-sharing, and led training on time series forecasting, embeddings, and RA

Founded and led the company's machine learning function, transforming R&D initiatives into production-grade predictive models that drove health care risk stratification, cost containment, and early interventions. Developed and deployed predictive models for chronic disease detection and ER visit forecasting, leveraging healthcare claims data to support early interventions and reduce avoidable admissions. Built real-time ML inference pipelines on PySpark and AWS EMR, reducing latency by 70% and optimizing infrastructure costs for large-scale healthcare data applications. Engineered secure, distributed ETL workflows using S3, LZ4 compression, and AWS-KMS encryption, improving throughput by 20% and ensuring HIPAA-compliant data handling. Applied neural networks and representation learning (embeddings, EFA/CFA) to extract hidden signals from clinical datasets, enhancing model performance and interpretability. Led internal research initiatives exploring time series modeling, NLP, and cluste

Designed and automated a disaster recovery framework using Ansible and AWS KMS encryption, reducing recovery time by 40% and ensuring secure multi-region data resilience for production systems. Led the migration of ETL infrastructure from Hadoop1 to Hadoop2 on AWS EMR, improving parallelism, increasing throughput, and reducing cloud costs across large-scale data pipelines. Developed and optimized high-volume data ingestion pipelines in Python for heterogeneous healthcare datasets, improving daily processing throughput and accelerating downstream analytics workflows. Diagnosed and stabilized Elasticsearch cluster performance through index lifecycle policies and capacity planning, increasing system stability by 35% and reducing unplanned downtime. Redesigned and unified batch data processing workflows into a centralized dashboard UI, improving operational visibility and reducing manual triage effort by over 50%. Collaborated cross-functionally with DevOps, Security, and Cloud IT teams to