Vishal Kumar

Collaborated with cross-functional teams to address business challenges using innovative techniques in big data analytics.

• Developed solution to evaluate promotion channel effectiveness for medicine brands to optimize channel spend and increase sales using IPMM model for SWISS Geography.
• Pharmaceutical product portfolios optimization. It aims to enhance decision-making, improve ROI, and ensure long-term sustainability in competitive pharmaceutical industry through data-driven insights and risk mitigation via allocating appropriate budget to various brands in dynamic market landscape.
• Developed NBA (Next Best Action) solution for promoting pharma brand in HCP community by targeting right HCP through right channel sequence to optimize brand sales.
• Developed HTA (Health Technology Assessment) model for predicting likelihood of new drug getting approved from NICE (U.K medical body) and launch price based on clinical trial performance and existing medical therapies cost.
• Currently, working on cognitive insight generation tool for digital engagement between Sales Rep and HCP's. The objective is to generate insight to augment Sales Rep Training, Sales Rep Evaluation, Improvement, Know-how sharing, personalize engagement with HCP's and Staffing based on compatibility between Sales Rep & HCP's.
• Developed RAG based application using LLAMA & BERT to empower researchers to extract valuable insights from vast collections of older research papers.

SMART CONTRACT ANALYZER

Developed a Transformer-based Smart Contract Analyzer Assistant using PyTorch to streamline legal compliance. This tool assists financial analysts by:

• Automated Legal Clause Extraction: Utilizes fine-tuned BERT models to identify relevant legal obligations within contracts, eliminating manual review of lengthy documents.
• Improved Efficiency: Saves analysts significant time and effort by extracting key data based on their queries.
• Enhanced Accuracy: Leverages semi-supervised learning for high-precision legal obligation identification.

CREDIT RISK VAR PREDICTION

Developed a comprehensive analytical system to empower proactive credit risk management for the enterprise. This platform leverages various data science techniques to provide:

• Early Warning Default Detection: Utilizes time series predictions (SARIMA) to identify loans at potential risk of default based on financial exposure metrics.
• Outlier Detection: Employs algorithms like Normal Distribution, Isolation Forest, and DBSCAN to pinpoint anomalous activities within the business.
• Financial Metric Variance Explanation: Leveraged correlation analysis and supervised learning to explain variations in financial metrics like STRESS and Exposure on Clearing House investments.
• Scenario Analysis: Implemented a Bayes-based scenario analyzer for predicting potential outcomes based on business data.

This platform equips the business with actionable insights, allowing for proactive risk mitigation, improved decision-making, and deeper understanding of financial health.

HEDGE FUND PROCESS AUTOMATION :

Developed a novel solution to extract tabular data from financial PDF reports with high accuracy. Traditional ML/AI methods proved unsuitable for this task due to the unstructured nature of PDFs.

This project leveraged my expertise in statistics and mathematics to:

• Analyze Hidden Data Patterns: Conducted intensive studies to identify patterns within the seemingly unstructured PDF data.
• Develop Statistical Extraction Rules: Utilizing statistical principles, designed robust rules for accurate tabular data extraction.

This innovative solution addresses a critical challenge in the financial industry, eliminating the need for manual data entry and streamlining data integration into financial ecosystems.

Wealth Manager Analytical Platform

Developed a comprehensive business analytics platform using Python to extract valuable insights from historical client data. This platform empowers stakeholders with data-driven decision making through Big Data Analytics, Predictive & Classification Modelling and Advanced Data Visualization Techniques. Some of problem statements solved under this initiative are as follows :

• Identified factors influencing client churn, enabling targeted retention strategies.
• Segmented client base for tailored product recommendations based on behavior during: Fee changes, Regulatory announcements (e.g., Repo rate), Global events (e.g., BREXIT), New product launches
• A/B tested decisions to measure their impact.
• Analyzed client sentiment to improve grievance resolution.

• Contributed to development and ongoing enhancement of a Python-based Reference Data Management System (RDMS) at Morgan Stanley. This collaborative effort ensures seamless flow of clean, consolidated, and accurate static securities data throughout enterprise. The focus was on developing and maintaining Python & SQL scripts to automate data updates and ensure reference database adheres to industry standards, providing a reliable data source for downstream systems.

• Developed a fraud detection application for ING Americas using Python and Sybase. This application identified and generated reports of suspicious transactions on ING insured contracts, ensuring compliance with US Patriot Act. Application processed massive datasets to extract actionable insights that empowered management to mitigate fraudulent activity.