Eitan Turok Email & Phone Number

New York City Metropolitan Area

• Develop model for demand forecasting in the non-stationary setting that outperforms traditional forecasting methods such as Sample Average Approximation (SAA), Holt, Naive, etc.• Our model works by decomposing the demand into its trend, seasonality, and noise and then running SAA on the noise

• Develop DP-SMOTE, a privacy-preserving algorithm to generate synthetic data points for imbalanced machine learning tasks • Published in ACM International Conference in AI in Finance (ACM-ICAIF'22) • DP-SMOTE uses uniform hypersphere sampling (computational thermodynamics), k-nearest neighbors (machine learning), disjoint subsets of r-nearest neighbors graphs (graph theory), the exponential mechanism (differential privacy)

• Implemented Wang’s NeurIPS 2020 paper on Clustering via Uncoupled REgression (CURE) in python• Demonstrated that CURE 1) outperforms traditional algorithms on classifying elliptically distributed data and 2) performs poorly on data with a non-linear decision boundary, confirming the theoretical results of the paper

• Co-founded StudBud, a social platform that matches students into ideal study groups with a machine learning algorithm during Covid-19 (variation of k-nearest neighbor & graph partitioning). • Fall 2020: StudBud had 1300+ sign-ups at Columbia, was published in the school paper, and won the Innovation and Enhancement Award.

• Conducted NLP research to answer the question: if a piece of writing contains multiple dates, which date is the most important to the article? • Developed Naive Bayes classifier to predict the most important date in a Wikipedia article with accuracy of 71% and an F1-Score of 0.83. Independently generated and cleaned my own dataset.

Bachelor'S Degree, 3.98

Activities and Societies: Columbia Data Product Initiative, Columbia Rock Climbing Club, Competitive Programming Coursework: Calculus.

Physics, First Year

Education record

Educational Institution based in Jerusalem, Israel.

Education record

Took the maximum of four AP courses: AP US History, AB Calculus, AP C Physics, AP Computer Science.