Data Scientist specializing in experimentation, forecasting, causal inference, and decision systems
Connect with me on LinkedIn.
I am a data scientist focused on forecasting, experimentation, causal inference, and decision systems. At Paramark, I help enterprise organizations measure marketing effectiveness through statistical modeling, incrementality testing, and marketing mix modeling. My work spans data engineering, model development, experimentation design, and analytical product development.
Forecasting, Predictive Modeling, Classification, Decision Systems
Developed forecasting and classification systems to improve sales pipeline visibility and revenue planning. Built and evaluated time-series forecasting models (SARIMA, Holt-Winters, Prophet) alongside opportunity scoring models using Elastic Net Logistic Regression and Decision Trees. Identified key drivers of deal conversion and improved forecast accuracy through systematic model comparison, feature engineering, and validation.
Simulation Modeling, Statistical Analysis
Built a stochastic simulation framework to evaluate disease transmission under varying assumptions around population density, infection rates, and vaccine distribution strategies. Analyzed how intervention policies impact outbreak severity, resource utilization, and population outcomes through scenario analysis and Monte Carlo simulation techniques.
Time Series Forecasting, NLP, Analytics Engineering
Developed an analytics platform for exploring and forecasting consumer finance complaints from the Consumer Financial Protection Bureau dataset. Combined topic modeling, sentiment analysis, n-gram analysis, Prophet forecasting, and ARIMA models to identify emerging complaint trends and predict future complaint volumes.
NLP, Recommendation Systems, Embeddings, APIs
Built an NLP recommendation engine that maps free-form user descriptions to Pokémon team compositions using semantic similarity. Compared multiple text representation approaches including Bag-of-Words, TF-IDF, GloVe embeddings, and transformer-based models to generate personalized recommendations from unstructured inputs.
 Thumbnail.png)
Machine Learning, Model Evaluation, Deep Learning, Computer Vision
Evaluated the robustness of image classification systems under adversarial attack conditions using Fast Gradient Sign Method (FGSM). Benchmarked EfficientNet, Vision Transformers (DeiT), and custom CNN architectures to measure model degradation, failure modes, and vulnerability to adversarial perturbations when distinguishing AI-generated and natural images.
NLP, Text Analytics, R Shiny
Built an NLP-powered writing analysis application that evaluates sentiment, ambiguity, and narrative progression within written content. Combined sentiment analysis, text summarization, and linguistic feature extraction into an interactive R Shiny application for content review and communication improvement.
Classification, Decision Systems, Model Evaluation
Built an interactive decision-support tool that translates classification model performance into expected business outcomes at different deployment scales. Uses confusion matrix metrics to estimate false positive and false negative volumes in operational settings, helping stakeholders evaluate model risk and decision tradeoffs.
Python, SQL, R, Git
Pandas, NumPy, Scikit-Learn
Snowflake, Fivetran, AWS, Tableau
PyTorch, HuggingFace
Causal Inference, Experimentation
A/B Testing, Marketing Mix Modeling
Incrementality Testing, NLP
Bayesian Modeling, Synthetic Controls
Predictive Modeling
Time Series Forecasting
ETL Design, Data Validation, Data Quality
Process Automation, Marketing Analytics
Project Management
Cross-Functional Leadership