Maplewood, New Jersey, United States
19 yrs. of experience in • Applied Data Science / AI ( -- Credit Risk Modeling in CRE, Corporate Loans, Mortgages, Subprime lending - Banking/Fintech; -- Fraud Detection using Agentic AI - Car dealers, Add-ons, and Auto Loans; -- Sales Forecasting for Semiconductor Manufacturing; -- Ad Measurement/ Marketing Science for CPG (Consumer Packaged Goods); -- Pricing Optimization for Hotels; -- Causal Inference - Impact of Marketing and Price Promotions; -- Recommendation Systems for Chain Restaurants; -- Migrating ML models from SAS to Databricks; -- Optimizing Lead Conversion in Online Lending platform - Fintech; -- Building AI Agents, RAG based Chatbots, and Agentic AI systems; -- Training and evaluating AI models as SME (Economics, Finance, Math & Data Science); -- Churn Prediction in Mass Media; -- Forecasting Fill Rate - Healthcare; • Capital Structure Arbitrage/ Equity Derivatives Trading; • Market and Credit Risk Management, Model Risk Governance, Model Validation, Valuation, and Internal Audit. Academic background in Engineering, Quant Finance, Economics, Econometrics, Bayesian Statistics, Data Science/Machine Learning, Reinforcement Learning, AI (ANN, CNN, Computer Vision), Generative AI (LLMs, RAG), Agentic AI, and introductory level Quantum Mechanics. Languages used: Python, R, SQL, PySpark, Stata, SAS, Matlab Infrastructure used: Databricks, GCP, AWS, Microsoft Azure Platforms used: Tekambi(Decision Engine)
- Quant Analyst @Google: Designed evaluation datasets, prompts, annotations, and rubric-based benchmarks to assess finance-focused LLMs, improving reasoning quality, factual accuracy, and instruction adherence. (via Wipro) - Data Science @Toorak Capital Partners: Designed and Built a PD & LGD model for secondary market (rehab) mortgages. (via Augment Analytics) - Data Science @Acxiom: Helped the client migrate data and ML models to Databricks platform, leveraging Delta Lake, Unity Catalog, and MLFlow (via KPI Partners) - Data Science @Hilton: Designed and built a model to assess the impact of price promotions on customer demand in test hotel groups by comparing against synthetic control matching groups, using Difference in Difference estimation with LMER (via KPI Partners / Vedan Technologies) - Data Science @KPI Partners: - Designed and built a model in Databricks to predict the churn rate for customers using propensity score modeling (XGBoost), and segmented customers based on churn likelihood, and key features; Further used Quasi randomized modeling (Propensity Score matching) to draw causal inference; Supervised Computer Vision Modeling Projects around Object Detection and Classification; Supervised Gen AI projects - building Agents and chatbots - Data Scientist @Mercor: Helped train and evaluate AI Models as a Data Science expert. - Model Risk QA @Citi: Reviewed the Model Risk IA Validation strategies around Bank’s AI/ML models, and Wholesale Credit Models in response to MRAs and Consent Order issued by Fed and OCC. (via Robert Half) - Model Trainer @Snorkel AI: Helped teach and evaluate LLM (Stanford HELM) for mathematical problem solving at the graduate level. - Model Trainer @Outlier.AI: Helped train and evaluate LMM (Large Multimodal Model) as an Economics expert.
- Data Science @Ford Motors: • Designed, and built a Fraud Detection model for Dealer Fraud Identification (using Isolation Forest), and Auto loan origination leveraging click-stream data; • Further, Architected a multi-agent AI system utilizing Gemini and a ReAct orchestration framework to extract features, verify data, and autonomously route complex invoice fraud to human reviewers with a continuous feedback loop. - Data Science @Wendy’s: Designed, built, and deployed a scalable Recommendation System using Association Rule mining (Apriori Algorithm) and Content based filtering (based on item category and flavor), on Google Cloud Platform (GCP), to enhance cross selling opportunities
- Data Science @Lam Research: • Built AI Agents in Azure Foundry with custom OpenAPI tools, connecting python API wrapper to internal knowledge/search applications. • Designed and built robust forecasting models for demand and revenue prediction using classical and ML based time series techniques (using Nixtla’s TimeGPT), on Databricks platform.
Data & Decision Science, Credit Risk Underwriting / Risk based pricing.
• Utilized factor analysis to help clients identify and leverage creative elements and ad features within podcast and simulated advertisements, enhancing campaign effectiveness and audience engagement. • Applied Signal Processing (Fourier Transform and wavelet analysis) to identify latent consumer behavior signals from sales time-series and engagement data • Implemented ensemble methods (Random Forest, Gradient Boosting) to identify key features (demographic, behavioral, and psychographic) for market segmentation, providing clients with actionable insights to optimize targeting and improve marketing strategies. • Optimizing the AB testing models used to estimate Brand Impact of digital campaigns on different platforms (LinkedIn, Spotify, Google, Meta) • Researching and implementing model specification changes in the ‘Market Lift Mixed Effects Regression Model’ to improve statistical significance, and also to accommodate ad hoc client requests. • Researching and Enhancing the functioning of Market Lift Impressions Calculator, used to calculate minimum Impressions required by an Ad campaign (for CPG) to achieve statistically significant Sales Lift (measured by Synthetic Control Matching Methods)