About

I am a Data Scientist with a background in Chemical & Biomolecular Engineering from UCLA, specializing in the intersection of deep technical implementation and high-level business strategy.

Currently a Data Science Manager at Capital One, I lead the development of high-performance Python frameworks that empower enterprise-wide teams to aggregate metrics and derive actionable insights from massive datasets. My career has spanned critical roles for the U.S. Census Bureau's Economy-Wide Statistics Division and Deloitte’s Risk & Financial Advisory practice, consistently focusing on Natural Language Processing, Machine Learning, and Process Automation.

My engineering roots at UCLA instilled a "first-principles" approach to problem-solving, which I now apply to the world of Big Data. Since my 2018 pivot at NovellusDX, I’ve been obsessed with turning raw information into strategic assets. I'm currently living in the Washington, D.C. area, so I’m always looking to collaborate with fellow tech enthusiasts in the DMV region; please reach out and let's connect!

Core Tech Stack

Case Studies

• Performed exploratory data analysis (EDA) using Winsorization to handle outliers, BoxCox transformations for normalization, and heatmaps for correlation analysis. Isolated the most statistically significant features impacting life expectancy to inform predictive modeling.

  Factors Affecting Global Life Expectancy: Identifying key public health drivers in a complex global dataset
  
  

• Built a Natural Language Processing (NLP) pipeline utilizing a Bernoulli Naive-Bayes classification algorithm. Successfully categorized reviews as having positive or negative sentiment with high precision and recall.

  IMDB Sentiment Classification: Automating the detection of viewer sentiment in thousands of unstructured movie reviews
  
  

• Developed a comprehensive research proposal, rollout plan, and A/B test evaluation strategy based on domestic flight data. Provided a data-driven framework for airlines to validate revenue-generating business decisions.

  U.S. Domestic Flight Analytics: Determining how U.S. airlines can optimize operations to improve revenue
  
  

• Comparative analysis of four regression models: Ordinary Least Squares (OLS), Lasso, Ridge, and ElasticNet. Optimized the models to minimize error, identifying Lasso and Ridge as the most effective for preventing overfitting in high-dimensional data.

  Predictive Modeling House Prices: Accurately predicting residential sale prices modeled on over 70 data features
  
  

• Implemented Decision Tree and Random Forest classifiers, utilizing NLP for feature engineering. Demonstrated the superior accuracy of Random Forest ensembles in handling non-linear weather relationships.

  Weather Classification via Decision Trees: Predicting weather patterns using historical atmospheric data
  
  

• Engineered features from text using NLP and trained a Support Vector Machine (SVM) classifier. Established a model capable of predicting if a recipe would be highly rated based solely on its textual composition.

  Recipe Rating Prediction (NLP & SVM): Predicting user ratings from key terms and descriptive tags
  
  

• Compared the performance of a Multi-Layer Perceptron (MLP) neural network against Gradient Boosting machines. Determined that Gradient Boosting provided more robust results for this tabular dataset compared to the deep learning approach.

  Precipitation Classification via Neural Networks: Classifying types of precipitation using complex meteorological sensor data
  
  

Side Projects

Resume

Contact

• ajgoldsman@gmail.com
• Cell Phone
• Linkedin