•Built an XGBoost credit-risk scoring engine with SHAP interpretability and an AI assistant for retrieving historical underwriting explanations.
•Developed, trained, and integrated ML models across Margin Intelligence and LAP, including XGBoost risk scoring, XGBClassifier for credit approval, XGBRegressor for probability-of-default estimation, and SHAP TreeExplainer for model interpretability.
•Automated loan-underwriting cash-flow assessment with an OCR-powered bank-statement extractor that produces structured FOIR, DTI, and spending profiles.
•Engineered a reusable multi-format OCR module for document classification and field extraction across fintech onboarding workflows.
•Built an OCR comparison workflow to evaluate extraction quality across document-processing approaches and select reliable pipelines.
•Implemented a HITL compliance layer for low-confidence extraction review and PAN/Aadhaar authenticity verification using HMAC-SHA256.
•Integrated a GitHub SCM SDK/API to save and push encrypted workflow exports directly to repositories with branch and file-path support.
•Owned data ingestion across 8+ global manufacturing plants using Qlik Replicate and Apache Spark for real-time and batch pipelines.
•Migrated legacy Oracle infrastructure to a scalable data lake, rewriting complex queries for MySQL and rebuilding SSIS ETL pipelines without data loss.
•Deployed Databricks PySpark workflows for large-scale manufacturing sensor data, reducing pipeline latency and recurring job failures.
•Enforced schema standards and data quality across Databricks and Snowflake ingestion layers.