Federico Scivittaro Email & Phone Number

Dallas, Tx, Us

Dallas, Texas, United States

Northbrook, Illinois, United States

I have spearheaded the analytical process for several projects at Astellas that have contributed to the overall understanding of patients and products within oncology markets.In one such project, I strove to better understand gastric cancer patients and caregivers, their understanding of various aspects of their disease, and their overall sentiment by performing natural language processing on social media posts. I began by collecting 6,700 posts from Twitter, Facebook, Instagram, Reddit, and several forums using the Digimind social listening platform. I cleaned each post by tokenizing, removing stopwords, and lemmatizing them using the Python NLTK library. I then identified the polarity and intensity of emotion of each post by applying the Flair sentiment analysis model, which performed better in empirical testing than VADER. I engineered an unsupervised topic modeling algorithm to examine the body of posts and classify them into three generalized categories. I also created a series of visualizations to summarize our key findings, including word clouds. I evaluated the results to identify the best opportunities to engage with patients and improve their understanding of their patient journey. I presented all of these findings and recommendations to a committee of 20 key stakeholders.In another project, I created a model to predict when early-stage AML patients would relapse. I began by crafting SQL code to extract data for 100,000 AML patients from an AWS Redshift data warehouse containing 140 million patients. I then used R to engineer 270 potentially explanatory variables to represent all aspects of the AML patient journey, including demographics, comorbidities, treatment history, and medical testing results. I then trained an XGBoost model on the data, using L1 regularization to remove excess features. Finally, I performed hyperparameter tuning to optimize the model. In the end, I achieved 80% recall and an AUC of 0.97 for predicting relapse in AML patients.

Chicago, Il

I created several products at E15 that focus on creating scalable software-as-a-service products that leverage data science, modeling, and machine learning.In one such project, I developed a linear regression to model employee and visitor demand for food and beverage at hospitals for a given day and meal period. I then created an algorithm to randomly sample from billions of possible combinations of food service delivery assortments to find an optimal food court design capable of meeting the predicted demand for a given hospital while minimizing capital expenditures with a 99.99% success rate. Lastly, I completed the pipeline in Amazon Web Services by connecting an intake form to SageMaker using S3 and Lambda, allowing hospitals to fill out a request for this service and receive recommendations and supporting visuals in under 10 minutes.For another project, I provided staffing configurations to venues for sporting events that would maximize concessions revenue while minimizing labor costs. I sourced, examined, and cleaned data for over 100 features and interaction terms in order to predict daily concessions demand given a sporting event’s characteristics—day of the week, time of day, home and away team quality, rivalries, weather, and more. I trained and tuned an XGBoost tree model, achieving 85% accuracy, and then created a custom algorithm to staff an entire venue based on the model’s demand forecasts and venue concessions data. Lastly, I automated the product to re-fit the model, generate demand predictions, and produce optimized staffing configurations each day and connected the outputs to a client-facing Django web app. The product currently generates and updates 1,000 configurations each day for over twenty clients in the MLB, NBA, NFL, NHL, and MLS.

Philadelphia, Pennsylvania, United States

I worked on several different projects throughout my time with the Eagles involving data cleaning in Stata, data analysis in R, and data visualization using R Markdown and the ggplot library. I examined optimal team strategy in the NFL by studying the effects of 2018's new kickoff rules and used strategies such as analysis of categorical data within contingency tables to derive insights and form recommendations to pitch to team decision-makers. I also examined short-yardage success by play type and crafted logistic regression models to fit the data and allow for a framework to drive future decision-making by the team.I also taught myself the R Shiny package in order to build an interactive web app complete with user interface and CSS styling. The app allows the user to efficiently query, filter, and display player statistics in a flexible and aesthetically pleasing fashion.Other duties involved charting proprietary data, data wrangling in Stata, and using Excel to quickly and clearly present data.

Master Of Science - Ms, Computer Science

Bachelor Of Arts - Ba, Statistics

Activities and Societies: International Leadership Council: Finance Division, Trott Business Program (formerly UChicago Careers.