Role Overview
Design, develop, and deploy AI, Machine Learning, and Generative AI solutions that address business problems across the disability insurance value chain, including claims segmentation, fraud detection, risk scoring, duration modeling, outcome prediction, and next-best-action recommendations. Translate business, actuarial, and claims objectives into practical AI/ML use cases, analytical designs, and production-ready solution components.
Responsibilities
- Work across the full ML lifecycle, including problem framing, data exploration, feature engineering, model development, validation, deployment support, monitoring, retraining, and continuous improvement.
- Build and optimize supervised, unsupervised, deep learning, NLP, and Generative AI models using structured and unstructured data.
- Develop NLP and document intelligence pipelines for extracting, classifying, summarizing, and interpreting information from medical records and claim notes.
- Support Generative AI solution development, including prompt engineering, embeddings, Retrieval-Augmented Generation, vector search, and model evaluation.
- Apply robust model validation practices, including performance tuning, bias detection, explainability analysis, and stability testing.
- Collaborate with data engineers, data scientists, and business stakeholders to ensure AI/ML solutions are scalable, secure, and reliable.
- Contribute to MLOps practices such as experiment tracking, model versioning, and automated pipelines.
- Follow responsible AI principles, model governance standards, and data privacy requirements.
Requirements
- Strong hands-on proficiency in Python and common AI/ML libraries.
- Solid understanding of machine learning and statistical modeling techniques including classification, regression, and clustering.
- Hands-on experience with NLP, deep learning, and document analytics use cases.
- Practical exposure to Generative AI and Large Language Models, including RAG and vector databases.
Skills
- Python
- PyTorch
- TensorFlow
- LangChain
- Scikit-learn