Shreyansh User

I've built machine learning systems that automate financial reconciliation, detect transaction anomalies, and assess credit risk. My projects combine practical ML engineering with explainability, regulatory alignment, and measurable operational outcomes.

In my Reconciliation AI platform, I automated transaction matching to reduce manual work by 80% and reconciliation errors by 90%. Adaptive learning improved accuracy to 95%, while automated reporting and root-cause analysis reduced audit preparation time by 60%.

I developed a hybrid transaction anomaly detection system using seven ML and deep learning models, achieving 99.8% AUC and 96% fraud-detection precision. I also built a Basel II-compliant credit-risk pipeline across 800K+ loan records, including PD, LGD, and EAD models with SHAP explainability and PSI monitoring.

As a Machine Learning Extern at Internpe, I used NLI, NLP, and transformer models including BERT, RoBERTa, and DeBERTa to identify policy-practice gaps across 10,000+ government documents. I created a Streamlit dashboard for 500+ analysts and delivered over 85% classification accuracy.