Vikas Phadtare

Wholesale Authorization Model (WAM) – Credit Risk Analytics:

• Developed and implemented dealer credit risk scorecards to support underwriting and continuous monitoring of wholesale finance portfolios.
• Performed feature engineering, WOE binning, variable selection, logistic regression modeling, score scaling, and validation using SAS and FICO Model Builder.
• Conducted portfolio monitoring, score migration analysis, model performance assessment, and governance reviews.
• Executed monthly scoring and decision strategy processes supporting dealer review prioritization and risk management.
• Enabled automated credit decisioning through model-driven decision strategies, resulting in approximately 50% of dealer authorizations being processed without manual review.

Bust-Out Fraud Detection:

• Developed and monitored machine learning models for proactive identification of bust-out fraud risk.
• Engineered bureau, behavioral, and account-level predictive variables for fraud detection.
• Evaluated model performance using Precision, Recall, F1 Score, ROC, KS, and PR-AUC metrics.
• Supported deployment of daily fraud scoring processes used to prioritize investigations and high-risk account reviews.
• Contributed to approximately $1M in fraud loss prevention through improved identification of high-risk accounts.

Smart Solution Search & Smart Case Search

• Developed AI-powered semantic search applications to improve solution discovery and reduce dealer case escalations.
• Built contextual search capabilities using transformer embeddings (all-MiniLM-L6-v2), BM25 ranking, and search relevance optimization.
• Developed Databricks-based embedding pipelines to maintain and refresh enterprise knowledge assets.
• Enhanced search functionality through user feedback integration and continuous product improvements.

Repeat Sales Analytics:

• Developed customer matching and entity resolution algorithms to identify repeat customers from sales data lacking unique customer identifiers.
• Leveraged customer attributes including name, address, phone number, and demographic information to create unified customer profiles.
• Built Power BI dashboards and generated insights on repeat sales trends across dealers, products, states, and districts to identify growth opportunities and support customer retention strategies.

• Developed generalised nested logit models to forecast monthly vehicle sales across France.
  Estimated incentive effects on demand and choice across multiple vehicle trims.
• Optimised marketing budget allocation using profitability and retail sales targets for upcoming quarter.
• Ran scenario forecasts and business simulations to guide sales planning.
  Partnered with stakeholders to turn model insights into marketing and sales actions.

• Delivered end-to-end forecasting solutions that support supply chain planning decisions.
• Strengthened demand planning by building machine learning and statistical forecasting models.
• Raised forecast accuracy by 10% versus demand planners and 5% versus existing solutions.

• Performed patient analytics on large-scale healthcare claims for site feasibility checks.
• Analysed patient populations by disease, treatment pathways, medication use, and demographics.
• Developed regression-based predictive models to estimate patient pools with limited claims.
• Collaborated with stakeholders to shape data-driven recommendations for trial site selection.

• Develop forecasting models for call volumes at monthly, weekly, daily levels.
• Apply time-series methods including ARIMA, ARIMAX, moving averages, and smoothing.
  Translate forecast call volumes into staffing plans with AHT and capacity targets.