Sankalp Mehani

About Sankalp Mehani

Freddie Mac leverages advanced ML models, including XGBoost and Random Forest, to revolutionize fraud detection in insurance claims. By integrating structured and unstructured data, such as policy limits and adjuster notes, the team enhances fraud scoring accuracy and automates manual review processes using NLP pipelines. These efforts create scalable, real-time solutions deployed on AWS platforms like SageMaker and Lambda. With a Master\'s in Data Science from the University at Buffalo, contributions focus on enabling data-driven decision-making through supervised and unsupervised learning, clustering, and model interpretability tools. Motivated by a passion for solving complex problems, the goal is to develop impactful AI systems that streamline operations and drive business value in financial and insurance domains.

Experience

1. Machine Learning Engineer

   Freddie Mac

   Feb 2024 — Present

   Designed and deployed supervised learning models (XGBoost, Random Forest) to detect fraudulent insurance claims using structured/unstructured data. Integrated features like policy limits, past behaviors, and incident timelines. Achieved enhanced fraud scoring accuracy and early detection.• Utilized unsupervised clustering to segment claims by behavioral patterns and anomalies, integrating third-party injury data to enhance fraud risk stratification and improve proactive detection in low-signal fraud cases.• Developed NLP pipelines with sequence-based neural networks to extract intent and red flags from adjuster notes. Boosted fraud detection rates by automating manual review triggers. Streamlined claim processing workflows using AI-driven text analysis.• Built and productionized fraud models on AWS using SageMaker, Lambda, and REST APIs integrated into SmartClaim Suite. Enabled real-time fraud scoring and automated routing of suspicious claims. Improved decision latency and early-stage fraud triaging.• Created end-to-end data pipelines in Python and SQL, orchestrated with Airflow for automation and retraining. Visualized fraud trends, model metrics, and risk scores via Tableau and Power BI. Ensured transparency using SHAP values and monitored model drift bi-weekly.

Education

• Vellore Institute of Technology

  Bachelor of Technology - BTech

  2016 — 2020

• University at Buffalo

  Master of Science - MS

  2022 — 2024