Open to opportunities

Shivanshu Singh

@shivanshusingh

Message

Data Scientist improving credit risk, fraud detection, and document intelligence with ML.

United States

Message

What I'm looking for

I’m looking for a role where I can build scalable, explainable ML for credit risk, fraud detection, and compliant document workflows—partnering to deliver real-time analytics and production-ready MLOps that improve latency, throughput, and decision quality.

I’m a Data Scientist with 3 years of experience in credit risk modeling, fraud detection, and document intelligence across BFSI datasets. I build end-to-end ML systems that balance accuracy, interpretability, and operational speed.

At Tungsten Automation, I developed credit risk prediction models using XGBoost, Scikit-learn, and SHAP, improving risk classification accuracy by 22%. I built fraud detection pipelines with PySpark, Apache Kafka, and Hadoop, enabling real-time transaction monitoring and reducing fraudulent transactions by 28%.

I also operationalize NLP and segmentation workflows for compliance and better decisions—automating KYC/AML extraction with BERT and LayoutLM to reduce manual verification effort by 40%. Using Kubeflow, Kubernetes, GitHub Actions, and MLOps practices, I’ve improved deployment efficiency by 35% while cutting model latency by 30% and scaling pipeline throughput by 2x.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist

Current

Tungsten Automation

Jul 2025 - Present (9 months)

Developed end-to-end credit risk prediction models using XGBoost and SHAP, improving risk classification accuracy by 22% across BFSI portfolios. Built real-time fraud detection pipelines with PySpark and Kafka (28% fewer fraudulent transactions) and implemented BERT/LayoutLM document intelligence for KYC/AML, reducing manual verification effort by 40%.

XGBoost scikit learn SHAP Pyspark Apache Kafka Hadoop BERT LayoutLM Kubeflow Kubernetes

AI/ML Analyst

Birlasoft

May 2021 - Jul 2023 (2 years 2 months)

Performed EDA and feature engineering on 5M+ financial records to build credit risk scoring models, improving prediction accuracy by 22% (AUC 0.87). Engineered ETL workflows and deployed ML pipelines with SSIS/Azure Data Factory and MLflow, reducing latency by 35% and deployment time by 40%.