Driving smarter audit reviews through NLP and ML in Financial Crime Risk, streamlining compliance and enhancing decision intelligence.

Skillset used: Text Preprocessing, TF-IDF, Feature Engineering, Logistic Regression, Power BI, Rule-based Labeling, ML Model Evaluation, Data Cleaning

🔍 What I did

• Spearheaded 4+ audit investigations in HSBC’s Financial Crime Risk (FCR) team, applying analytical rigor to 250+ cases to ensure effective and compliant risk controls.
• Developed Power BI dashboards for real-time audit tracking and reporting, reducing turnaround time by 20% and enhancing stakeholder visibility.
• Initiated a machine learning-based narrative evaluation pipeline, automating quality checks on 5,000+ FCR review cases:
  • Utilized TF-IDF for text vectorization and structured input for modeling.
  • Performed advanced feature engineering:
    • Cleaned inconsistent text formats.
    • Merged fragmented narratives for contextual integrity.
    • Filtered out system-generated comments to reduce noise and boost precision by 15%.
• Implemented a rule-based labeling system with a 0.85 confidence threshold, flagging low-confidence cases for manual review to ensure high audit accuracy.
• Evaluated multiple ML models (Logistic Regression, Decision Tree, Random Forest, XGBoost) and deployed Logistic Regression (91% accuracy) for its balance of performance, interpretability, and generalizability.

📈 Impact & Insights

• Process Optimization: Automated audit checks, significantly reducing manual review effort and improving narrative consistency.
• Risk Reduction: Enhanced the accuracy of financial crime case reviews, leading to better compliance with regulatory standards.
• Time Efficiency: Delivered insights and dashboards that empowered teams to make quicker, data-backed decisions.
• Data-Driven Judgement: Equipped audit teams with a scalable tool to assess narrative quality, boosting clarity and actionability.

🚀 Learning Outcomes

• Gained hands-on experience in real-world NLP for compliance and risk analytics.
• Strengthened ability to design interpretable ML pipelines tailored for high-stakes domains.
• Learned to balance automation with human oversight in risk-sensitive processes.
• Sharpened skills in Power BI storytelling to communicate complex audit insights effectively.