Lead AI Scientist | Machine Learning Modeling | Pricing & Risk & Marketing & Operation & Servicing Strategy | AI/ML Engineering
I am a Lead AI Scientist with 19+ years of experience in AI-powered recommendation engines, LLM/RAG, pricing optimization, credit scoring, loss forecasting, fraud detection, and building large-scale ML pipelines. My expertise lies in predictive modeling, model validation, and AI-driven business analytics for fintech, banking, payment, e-commerce, real estate, investment, and platform industries. Proven leadership 12+ years in leading cross-functional teams in implementing enterprise-wide ML solutions, managed and mentored data scientists, and spearheaded strategic data initiatives resulting in significant operational improvements.
This is my personal portfolio site. The sample projects here are prototypes, MVPs, and toy projects created to demonstrate my skill set during my master’s program and in my spare time. They are not work-related and contain only my own proprietary work. Data sources are primarily from Kaggle, arXiv, and various open-source APIs.
If you have any questions or are interested in collaboration, feel free to reach out.
- LLM RAG based AI ChatBot: A powerful AI ChatBot App that lets you chat with multiple documents using LLM and RAG.
- MLOps using Docker & Kubernetes: End-to-end ML pipeline, MLflow, CI/CD implementation using Docker containers and Kubernetes orchestration for scalable model deployment.
- House Price Prediction using Deep Learning PyTorch and ML Pipeline: A deep learning model built with PyTorch to predict house prices based on real estate features.
- ML Pipeline for On-premise and AWS cloud: A complete on-premise/AWS cloud machine learning pipeline setup for training, validation, and deployment with local servers, AWS EC2/S3.
- Data Pipeline using Airflow: End-to-End Data Pipeline using Spark, S3, Databricks, and Airflow: Word Count, User Behavior Analysis, UDF-based Segmentation, and Daily Workflow Automation.
- Real-time Crypto Price Analytics Platfor: Built a real-time BTC/ETH price tracking system using Kafka, Spark Streaming, Redshift, Airflow, and Stremlit, featuring live dashboards, Slack alerts, and volatility trend analysis.
- Credit Risk Scoring: Creating risk-based underwriting scorecards leveraging logistic regression, XGBoost, and deep learning models.
- Loss Forecasting: Developing robust models to predict credit losses (PD/LGD/EAD) using macroeconomic factors, delinquency trends, and time-series analysis.
- Fraud Detection: Implementing anomaly detection models using unsupervised learning (Isolation Forest, Autoencoders) and supervised learning (CatBoost, XGBoost).
- Residual Value Modeling: Forecasting vehicle residual values using econometric models like PROC MIXED and machine learning approaches.
- Pricing Optimization: Optimizing pricing strategies using reinforcement learning, JD Power competitive rate analysis, and market elasticity modeling.
- Interest Rate Forecasting: Predicting interest rate trends through time-series forecasting (ARIMA, LSTM) and economic indicators.
- Behavioral Score (Collection Model): Developing early-stage delinquency prediction models for efficient collection strategies.
- Churn Prediction (Retention): Building ML models to identify customers at risk of attrition, utilizing survival analysis and XGBoost.
- Propensity (Return to Market) Model: Developing models to predict customer likelihood of returning to market based on behavioral signals and transaction data.
- Customer Satisfaction (Sentiment Score) Model: Using NLP techniques (BERT, LLMs) to extract sentiment insights from customer reviews, survey responses, and service interactions.
- Marketing Mix Modeling (MMM): Analyzing the effectiveness of various marketing channels through econometric regression and Bayesian inference.
- Languages: Python (Pandas, NumPy, Scikit-learn, PyTorch, TensorFlow), SQL (BigQuery, Snowflake, Redshift, MySQL, PostgreSQL), R, SAS, Java, Scala, NoSQL (Mongo, Couchbase), React, Node/Next.js
- Big Data & Cloud: Databricks, DBT, Apache Spark (PySpark, Spark ML, MLlib), Apache Beam/Flink, AWS (S3, Lambda, EC2), Azure, GCP (BigQuery, Vertex AI), OCI, Airflow DAG orchestration
- Machine Learning & Framework: XGBoost, LightGBM, CatBoost, Logistic Regression, Random Forest, Neural Networks (Transformer, LSTM), H20, AutoML, TensorFlow, PyTorch, Scikit-lear, Generative AI-LLM (Langchain, GANs, VAEs, Langroid), NLP (NLTK, Vader), Unsupervised learning (Isolation Forest, Autoencoders)
- Visualization & Reporting: Tableau, Power BI, Looker, D3.js, Matplotlib, Seaborn, Databricks Quick insights, AWS QuickSight
- MLOps & Deployment: Docker, Kubernetes, MLflow, Spark + Airflow, GitHub Actions, API Deployment (FastAPI, Flask), Git, Bitbucket, AWS S3, scalable DAG setup
- Data Warehousing & Data Lake: Oracle Object Storage, Apache Hadoop, Apache Hive, Amazon Redshift, Snowflake
- Pipeline & Automation: Apache Airflow (SparkSubmitOperator), daily batch ETL scheduling, parameterized DAGs, S3 integration
- Extensible Architecture: DBT / Snowflake integration, containerization (Docker), and CI/CD (GitHub Actions) ready with minimal refactor
- LLM RAG Based GenAI ChatBot – A powerful AI ChatBot App that lets you chat with multiple documents using LLM and RAG.
- LLM RAG PDF AI Platform – PDF files analytics LLM RAG platform using LangChain, Hugginface, and Strealit.
- Advertisement Analytics Model – Designed Advertisement Analytics Marketing Model for Customer Segmentation.
- Recommendation Model – Designed and implemented recommendation models such as book, movie, etc.
- Fraud Anomaly Detection – Built an AI-powered fraud detection model that reduced fraud by 25%.
- Churn Prediction Model – Developed a churn prediction model that enhanced retention rate by 18%.
- Credit Scorecard Model – Developed customer behavioral credit scoring models to mitigate the risk.
- Loss Forecasting Model – Built loss forecasting models compliance with CECL/ALLL.
- Residual Value Model – Developed an RV model using PROC MIXED to enhance auto finance predictions.
- Developed a fraud anomaly detection model that reduced fraudulent transactions by 25%.
- Designed a credit scorecard model that improved underwriting efficiency, leading to a $15M ROA increase.
- Led the residual value modeling project, accurately forecasting used car values and optimizing lease pricing.
- Built an AI-powered customer sentiment model, reducing customer complaints by 18%.
- Delivered an interest rate optimization framework, enhancing portfolio returns with predictive modeling.
- Innovative Applications of Artificial Intelligence (IAAI) Award (2023)
- Supported and oversaw a ML deep learning-based automobile market price forecasting model, enhancing risk assessment and transaction control.
- Finance Journal Recognition (2013)
- Featured for best practices in loss forecasting models and enterprise risk management systems (ERMS).
- Researching LLMs and GenAI applications in risk & marketing modeling.
- Exploring causal inference techniques for better decision-making in financial products.
- Implementing automated MLOps pipelines for scalable credit risk modeling.
Let's connect! If you're interested in innovative AI applications and ML modeling for measurable benefits, feel free to reach out!