Gayathri Harihara Subramanian

Client's Profile: Leading Ride Sharing Organization in North America (Risk & Fraud Analyst)

• Led a team of analysts within LatentView, overseeing end-to-end fraud detection initiatives, ensuring timely delivery, cross-functional collaboration, and stakeholder alignment
• Developed Haversine geo-validation to toll reimbursement fraud, leveraging Hive SQL and Python for data processing, leading to annually savings of $2.1M
• Detected and mitigated rider-driver tip collusion by analyzing tip-to-fare ratios using statistical outlier detection methods, and implemented API-level rule-based filters to enforce fraud controls
• Created rider churn analysis visualizations in Looker Studio by connecting Apache Hive via Simba ODBC driver, enabling insights into retention trends, usage frequency, and cohort behavior.
• Implemented anomaly detection techniques—such as model score monitoring, drift detection, and threshold-based alerting—to flag degradation in model precision and trigger early investigation of performance issues in production models

• Built a Streamlit-based internal training and assessment app for SQL and Python, deployed on Azure App Services with Cosmos DB for metadata storage and Azure Blob Storage for handling files.

Client's Profile: global e-commerce giant (Tax and Regulatory Analyst)

• Built an algorithm using SparkSQL to classify transactions impacted by VAT regulation changes across European and UK markets, determining whether amounts should be recouped or reimbursed
• Leading analytics initiatives for regulatory changes including 1099-K, DAC7, EEK, DSA and VAT updates, working with legal and compliance teams to scope business impact, identify affected transaction flows, and monitor go-to-market readiness via Tableau dashboards powered by SparkSQL

Client's Profile: leading ride-sharing platform in North America (Risk & Fraud Analyst)

• Card Testing Attack Mitigation: Detected high-velocity, bot-driven card testing attacks by implementing entropy scoring, card fingerprinting, and analyzing authorization failure patterns. Deployed real-time detection using NOSTRIL and custom regex filters to identify gibberish inputs while preserving user experience. Built detection logic using Hive SQL and Python, with reporting via Mode Analytics
• Loss-Friction Analysis & Rule Engine: Developed a precision-driven rule engine using internal and third-party risk signals to detect high-risk transaction patterns with minimal false positives. Migrated existing Python models to R to improve prediction accuracy for chargeback and debt rates. Utilized Hive SQL, Hive Python, and Mode Analytics for rule evaluation and monitoring

• Analyzed A/B test and product usage data using R and Excel to generate actionable insights that informed product refinement and feature prioritization
• Supported strategic decision-making by analyzing customer survey and market research data to identify key user segments and drive customer-focused improvements
• Contributed to cross-functional team discussions by sharing insights, addressing analytical challenges, and collaborating on data-driven solutions

• Built energy loss estimation model using linear regression and splines in R, with visualizations in Spotfire, and achieved third-party certification
• Analyzed historical failure data using SQL Server and cube data to build correlation models, enabling proactive service planning and reducing unplanned downtime through early identification of failure patterns
• Performed component reliability analysis in R to calculate failure rates and support maintenance strategy decisions

• Developed a power-energy distribution analysis model for transmission clutches using Python, Teradata, and Oracle SQL, including data cleaning, computation, and Tableau visualization, leading to software and material design changes
• Supported migration of legacy VBA macro-based code to Python, improving maintainability and reducing processing time and operational cost
• Conducted root cause analysis of bearing failures using Python, uncovering new insights that guided hardware design improvements by product engineering teams