About Me

Hello! I am Diego Escobedo, and I recently graduadated from MIT with a BS in Computer Science and Engineering. I am deeply interested in machine learning, specifically its applications to medical research and financial markets. In my free time, I love DJing, messing around on my film camera, and playing basketball (poorly). Please don't hesitate to email diegoescobedo ατ mit døt edu if you have any questions, ideas, or just want to chat. Hope you enjoy my website!

Education:
  • Bachelor of Science in Computer Science and Engineering - MIT

    Class of 2022

    GPA: 4.7  |  Major GPA: 4.9

    Clubs: Phi Delta Theta, Amphibious Achievement, DanceTroupe, College Diabetes Network

Download Resume

Coursework

Computer Science
  • 6.046
    Design + Analysis of Algorithms
  • 6.031
    Software Construction
  • 6.033
    Computer Systems Engineering
  • 6.004
    Computation Structures
  • 6.864
    Advanced NLP (G)
  • 6.867
    Machine Learning (G)
  • 6.869
    Advances in Computer Vision (G)
  • 6.884
    Seminar in RL + NLP (G)
Mathematics
  • 18.02
    Multivariable Calculus
  • 18.062
    Math for Computer Science
  • 18.06
    Linear Algebra
  • 18.600
    Probability + Random Variables
  • 6.041A
    Probabilistic Systems Analysis I
Economics
  • 14.01
    Microeconomics
  • 14.02
    Macroeconomics
  • 15.815
    Applied Behavioral Economics
  • 15.053
    Optimization Methods
  • 15.911
    Entrepreneurial Strategy
  • 14.41
    Public Finance and Policy

Languages & Tools

Strong Knowledge
  • Python
  • SQL
  • Git
  • Pytorch
  • Java
Proficient
  • Google Cloud Platform
  • JavaScript
  • HTML / CSS
  • Julia
  • TensorFlow / TensorFlow.js

Work Experience

Industry :
  • Quantitative Researcher Intern - Citadel
    Tenure: Summer '22 Division: Data Strategies Group Tech Stack: Python | SQL
    • Onboarded and analyzed two novel datasets for the Consumer Staples team, generating alpha through major improvements in understanding store-level inventories and insight into retailer margins.
    • Developed a pricing algorithm to determine the true underlying dynamics of items affected by factors such as supply chain shortages and discounts, leading to significant accuracy gains when modelling CPI and other price-related metrics.
  • Quantitative Strategist Intern - Goldman Sachs
    Tenure: Summer '21 Division: Consumer & Wealth Management Tech Stack: Slang
    • Updated a performance attribution report generation system to include advanced tooling for benchmark analysis and a robust set of linking algorithms. New features used for tens of thousands of reports covering $1 trillion in AUS.
    • Created a tool to enable client teams to originate, price, and adjust custom fixed-rate interest products. Currently being used to roll out new bank loans to UHNW clients.
  • STEP Intern - Google
    Tenure: Summer '20 Division: Google Research Tech Stack: HTML/CSS | JavaScript | Python | Java | Google AppEngine | GitHub | TensorFlow/TF.js
    • Created a fantasy basketball engine, where users could build a custom team, insert them into a real NBA season, and use a model to predict the outcome of a match between any two teams.
    • Abstracted players into ~30 efficiency and counting stats and integrated their identities into a “bag of players” feature, which allowed for the deep neural network to handle any match between any set of players.
    • Model performed at ~78% accuracy, better than most experts and scientific papers.
  • Global Analytics and Insights Intern - Electronic Arts
    Tenure: Summer '19 Division: Maxis Studios Tech Stack: Python | GitLab | HiveSQL | Hadoop | AWS Redshift
    • Leveraged data from ~1.5M players and developed a RF classifier to optimize targeted advertising and improve key business KPIs. Used data from first few hours of gameplay to predict spend outcome and send offers to ensure indecisive players are brought into the company’s pack buyer network.
    • Generated daily reports and produced ad-hoc analysis for a variety of business units and studio leadership, as well as a long-term project building out Maxis’ data infrastructure.
    • Performed at ~88% accuracy.
Research:
  • Undergraduate Researcher - MIT Computer Science & Articial Intelligence Laboratory
    Tenure: Fall '21 - Spring '22 Division: Geoemetric Data Processing Group // NeuralODE Project Tech Stack: Pytorch/Pytorch Lightning | torchdiffeq
    • Currently developing applications for the Neural ODE family of DNN models in the bioinformatics and graphics space.
    • Responsible for developing models that can predict the developmental time courses followed by stem cells during cell reprogramming, using scRNA-seq profiles.
  • Undergraduate Researcher - MIT Media Lab
    Tenure: Winter '18 Division: Viral Communications Group // Layer Project Tech Stack: Python | GCP (Compute Engine) | Wikibase
    • Developed a content-based recommendation algorithm that maintained privacy by performing binary representation matrix calculations on end devices.
    • Built detailed and robust user profiles by implementing a naïve Bayes classifier with a multivariate Bernoulli event model.
    • Media Lab project page can be found here. Unfortunately, the page was removed, but the web archive is still around.
  • Multimodality Imaging Lab Intern - Stanford School of Medicine
    Tenure: Summer '17 - Spring '18 Division: Molecular Imaging Program at Stanford Tech Stack: Python | Raspbian | Arduino
    • Invented a ‘smart toilet’ platform that analyzes bodily fluids to enable the early detection of diseases such as diabetes, urinary tract infections, and STIs, by collecting and matching biometric and medical data to create a longitudinal profile of patients’ health.
    • Used MATLAB to create an image acquisition and segmentation algorithm that detects areas of interest in urinalysis assays and translates RGB color profiles into biochemical data.
    • Links:
    Early prototype:

Projects

Applications:
  • Retweetable

    Final project for 6.864, MIT's class called "Advanced Natural Language Processing". I made a Variational Autoencoder with byte-level byte-pair encoding and nucleus sampling decoding to generate funny tweets. Trained on a custom dataset created by scraping Gen-Z Twitter's funniest tweets.

    View a static version here, including all implementation details.

  • ball.ai

    Capstone project for my STEP Internship. Along with two other interns, we created a fantasy basketball engine. The idea behiond it wasd to be able to create a "fantasy" team with players from any era and any team. My part of the project involved creating a neural network that could effectively predict the outcome of games WITHOUT using any team-level data (since we only have player-level data about these fantasy teams). I succesfully managed to create a model that worked, and in fact beat a few scientific papers (detailed in this meta study ). Created the model in Tensorflow and then deployed to TensorFlow.js on Google App Engine.

    View the website here. I made the landing page and the "Simulate" page, the rest were created by other team members.

  • COVID-19 Walkthrough

    Final project for 6.S083, MIT's class called "Computational Thinking in Julia". I made a walkthrough on how to use gradient descent to fit real-world COVID data to a SEIR Model. Fitting this data is extremely important, as these models give us powerful insights about the spread of the disease, but finding the right parameters for the differential equations is not an easy task.

    View a static version here.

  • Hypermines

    N-dimensional minesweeper. Probably not as good-looking as Microsoft's version, but I don't think you can play their version in a hypercube. Made for 6.009, MIT's Fundamentals of Programming course. Web app not available, but feel free to download the source code!

    a game of hypermines

    To run the app from your computer, unzip the files, right click server.py and go to properties, and copy the location. Then, open up the terminal, and type "cd" and then paste the location of server.py. Then, type "python3 server.py", wait a little bit, and then type "localhost:8000" on your web browser.

  • Zoo Mania

    Inspired by the timeless Bloons Tower Defense, I made a small tower defense game. It's a little slow and buggy, but it works alright. Made for 6.009, MIT's Fundamentals of Programming course. Web app not available, but feel free to download the source code!

    a game of zoo mania

    To run the app from your computer, unzip the files, right click server.py and go to properties, and copy the location. Then, open up the terminal, and type "cd" and then paste the location of server.py. Then, type "python3 server.py", wait a little bit, and then type "localhost:8000" on your web browser.

Open Source Contributions:
  • Basketball Reference Scraper

    Had a lot of fun working on my very first open-source project, a webscraper for the website Basketball Reference. The repo for the scraper can be found here. As you can see from my pull requests (#5 and #14) and my raised issue, I was actively working towards improving this project and managed to find a serious bug. I used this scraper when I was first learning how to use ML techniques to analyze NBA data, and it has been extremely helpful since then.

More About Me

Volunteering and Leadership Experience:
  • Phi Delta Theta - Massachussetts Gamma Chapter
    Tenure: Fall '18 - Spring '22 Roles: President, Recruitment Chair, Social Chair, Scholarship Chair
    • As President, redesigned the bylaws, improved our safety procedures, and spearheaded efforts for a house renovation. In charge of coordinating over 20 officers’ efforts in a variety of areas, including house management, social, and academic endeavors.
    • As Recruitment Chair, managed a budget of 25,000 dollars for a variety of events over Campus Preview Weekend and Rush Week. Coordinated logistics for over 30 events, ranging from small meet-and-greets to house tours attracting hundreds of people.
    • Webpage can be found here.
    Photos of a hiking retreat in Vermont, Faculty Dinner with Prof. Jason Ku, and our famous marble staircase.
  • Amphibious Achievement
    Tenure: Fall '18 - Spring '20 Roles: Mentor, Director - External Relations Branch
    • Coordinated and continued developing a dual athletic-academic mentorship program for over 70 high school students with the purpose of expanding their higher education opportunities.
    • Helped students develop the confidence needed to achieve their goals by implementing a wide-ranging, hands-on curriculum intended to promote a love of learning.
    • Webpage can be found here.
    Photos of both our athletic and our academic sessions. We love Sundays!
  • MIT Interfraternity Council
    Tenure: Winter '19 - Winter '20 Role: Vice President
    • Established and led the Constitution Review Committee. Rewrote sections to revamp internal judicial system and more clearly delineate powers and responsibilities.
    • Part of the COVID-19 Response Team. Planned and executed programs to get students safely off campus during a time of crisis while ensuring well-being of constituents and houses.
    • Created the Diversity Committee, in order to ensure the IFC’s commitment to nondiscrimination and celebrate the diversity of one of MIT’s largest student organizations.
    • Webpage can be found here.
    Photos of the IFC Executive Board at the NGLA conference + a nice headshot.
  • MIT AppInventor - Brazil
    Tenure: Winter '20 Role: Teacher
    • Spent a month in Brazil teaching small businesses and young students how to use AppInventor, an MIT initiative that seeks to democratize app development
    • Designed a curriculum that would help small business owners take charge of app development and bring the Brazilian informal sector to the 21st century.
    • Blog post about the experience.
    Photos of the teaching team for AppInventor and an action shot showing a student how to use the tool.
Hobbies:
  • DJing
    Don't have any mixes ready for the spotlight, but here's some of my spotify playlists for inspo:
  • Photography
    A nice little compilation I made: