Shashank P.

About Shashank P.

AI / ML Engineer with 4+ years of experience designing, building, and deploying production-grade machine learning, deep learning, and Generative AI solutions across healthcare and financial services domains, with strong expertise in end-to-end ML pipelines including data preprocessing, feature engineering, model training, evaluation, deployment, monitoring, drift detection, and automated retraining. Hands-on experience in NLP, computer vision, semantic search, and Retrieval-Augmented Generation (RAG), leveraging Large Language Models (LLMs), dense embeddings, vector databases, and advanced prompt engineering techniques such as few-shot and instruction-based prompting. Proficient in building scalable and explainable AI systems using PyTorch, TensorFlow, scikit-learn, LangChain, and Hugging Face, exposing models through RESTful APIs with FastAPI and Flask, and deploying containerized services using Docker, Kubernetes, and CI/CD pipelines. Strong background in explainable AI using SHAP and LIME to support clinical review, regulatory compliance, and risk governance, with proven ability to deliver secure, low-latency, and audit-ready ML solutions. Demonstrated success collaborating with clinical, risk, compliance, data engineering, and business stakeholders to translate complex model outputs into actionable decision-support workflows and reliable human-in-the-loop AI systems in highly regulated environments.

Experience

1. Ai Ml Engineer

   CVS Health

   Aug 2024 — Present · US

   Built ML pipelines on EHR and claims data, performing data preprocessing, feature engineering, and validation using Python, Pandas, SQL, and scikit-learn.• Developed risk stratification and healthcare utilization models using XGBoost, Random Forest, and PyTorch, supporting population health initiatives.• Designed NLP pipelines for clinical text processing using BERT, SpaCy, Hugging Face, and NLTK.• Implemented Retrieval Augmented Generation (RAG) using LLMs, embeddings, vector databases, and LangChain for grounded clinical knowledge retrieval.• Applied prompt engineering techniques (few-shot, instruction-based prompts) to improve LLM output accuracy and consistency.• Built semantic search systems using dense embeddings and vector indexing.• Developed CNN-based computer vision models using PyTorch, TensorFlow, TorchVision, and OpenCV for medical image and document processing.• Evaluated models using Precision, Recall, F1-score, ROC-AUC, BLEU, and ROUGE, prioritizing recall for safety-critical use cases.• Applied Explainable AI techniques using SHAP and LIME to support clinical review and compliance requirements.• Exposed models through RESTful APIs built with FastAPI and Flask, enabling real-time inference.• Containerized and deployed services using Docker, following CI/CD pipelines for controlled releases. Implemented model monitoring, drift detection, and retraining workflows in production environments.• Ensured HIPAA-compliant data handling, secure access controls, and audit-ready documentation. Collaborated with clinical, data engineering, and compliance teams to align solutions with healthcare workflows.• Designed decision-support workflows combining ML predictions with clinical rules to assist care teams in risk assessment.• Implemented human-in-the-loop feedback mechanisms, allowing clinician review to improve model reliability over time.

Education

• University at Buffalo

  Masters , Data Science