Software Engineering

> Alfonso.origin
  => "Monterrey, Mexico"

  => "Minerva Schools at KGI Class of 2022"

> Alfonso.major
  => "Computational Sciences and Physics"

> Alfonso.hobbies
  => ["Latin Rhythms", "Soccer", "Foreign Languages", "Food", "Reading", "Traveling"]

  => Latest Work Experience
        company: ""
        role: "Software Engineering - Computer Vision"
        startDate: "January 2020"
        endDate: "May 2020"

> Alfonso.languages
  => {"Spanish": "Native", "English": "Fluent", "Portuguese": "Fluent", "French": "Proficient"}

> Alfonso.resume
  => "asantacruz_resume.pdf"

  => ""


MiPay Payroll Platform   Designed, proposed and developed a web application for the Work Study program at Minerva Schools at KGI. Students had complained in the past about not having an interface with all their financial information handy (worked hours, payments received, remaining hours to work as part of the scholarship, among others). This web application allows them to report their hours, consult schedules and budget their income. For managers, the app displays all their interns in one place as well as interns' submissions and payments records. This allows managers to administer the amount of work they give to their interns. Last, the platform uses Google Sheets API to integrate with the current procedures of the company. Specific teams can add or remove workers, update managers and positions automatically, etc. Automatic reminders and pay period closures are set using Time Based CLoud Triggers. Utilizes Node.js, Express.js, Mondo DB, JQuery, Bootstrap, HTML, CSS, Google Sheets API, Google Apps Script, Gmail API, Heroku, Docker and Google Cloud Functions.

Automated Offer Letter Generator   Deployed using GCP (Cloud Functions, Storage, Apps Scripts and Stackdriver), leverages Google Sheets API, Google Drive API, Gmail API and HelloSign API to automate a pipeline to generate, send and save (when signed) offer letters based on authorization conditions and data input from many company teams on an internal database. Nearly 60 hours of work were saved during Summer 2020 as a consequence of this computer program. Crushed the average overall time to process an offer letter manually from 15 min. to a few seconds. Last, 250+ offer letters were processed this Summer.   Designed and developed an image captioning pipeline using a combination of a Convolutional and a Recurrent Neural Network on Bahdanau Attention with Imagenet weights, using the IU X-Ray Dataset. Created Tensor Images through InceptionV3 architecture. Imported an Encoder CNN and a Decoder RNN from Keras. The model surpassed validation accuracy metrics (BLEU, ROGUE AND CIDER) by 3% with respect to academic papers. See project on site

Diabetic Retinopathy Detector   Implemented progressive resizing in a Res-Net 34 CNN to develop an image classifier (~92% accuracy) from 3000+ images of eye retinas in fast/ai and Pytorch to speed up diabetic retinopathy detection in rural areas of India.

BudgetApp   Developed a Web App to help my peers manage their personal finances during our college experience. Utilizes Google Sheets API in NodeJS to export and import statements from sheet formats. Ensures data privacy through Google OAuth2. Uses HTML/CSS and JS for the Front-End.

Web Image Crawler in Selenium   Image Crawler developed using Selenium Web Crawler in Python. Very useful and handy solution to crawl images for Computer Vision (Image Classification, Segmentation, Image Captioning) projects when datasets are not publicly available. Designed in such a way the it accepts a .txt file with the names of the items to be crawled in Google. The app then opens a browser window, scraps all the links of the images obtained from the Google search of each item and downloads the images to your local machine. Last, it stores the images in folders ready to be processed into a Data Load while also suggesting solutions for data imbalance, if any.

Zenefits Terminator   My Work-Study manager asked me to update 700+ employees from Zenefits Payroll App. After thorough research and chat with tech support, I realized the platform lacked bulk operations for my specific request. Thus, I designed and implemented a data pipeline and bot to automate the updates using Google Sheets API and Selenium Web-Driver in Python. The program decreased the expected time of each update from 7 to less than 1 minute per employee in the payroll database without the need of any human intervention.


Minerva Schools at KGI September 2018 - May 2022
Computer Science, B.S. (GPA: 3.7) San Francisco, CA

Relevant Coursework:

  • CS162: Software Development. Building Powerful Applications
  • CS110: Computation: Solving Problems with Algorithms
  • CS112: Information Based Decisions. (Data Science using R)
  • NS110U: Physics of the Universe
  • CS111A: Continous Methematical Systems (Multivariable Calculus)
  • CS111B: Linear Mathematical Systems (Linear Algebra)

Programming Languages
Python 3 Javascript ES6 R HTML CSS Dart

Web Frameworks
Node.js Express.js JQuery Bootstrap Flask


Google Cloud Platform Heroku Docker Travis CI Git Flutter

Important Libraries
Tensorflow Keras Sagemath Numpy Stack Google APIs Selenium

Minerva Project Septemeber 2018 - Present
Finance and Human Resources Intern San Francisco, CA

- Reduced Finance Team’s working time by 20 hrs/wk through developing and implementing a web payroll platform to track salaries and working hours of 500+ employees.
- Developed an automated pipeline for the HR department to create, send and store 250+ contracts, reducing 15 min. of work per hire.

Python 3Javascript ES6HTMLCSSNode.jsExpress.jsMongoDBHerokuGoogle Apps ScriptGoogle Sheets APIGoogle Drive APIHello Sign APIBoostrapJQueryTravis CIGit Spring 2020
Software Engineer Intern - Computer Vision Hyderabad, India

- Designed and developed an image captioning pipeline by applying a CNN-RNN in Bahdanau Attention with Imagenet weights, using the IU X-Ray Dataset and InceptionV3 architecture. The model surpassed validation accuracy metrics (BLEU, ROGUE AND CIDER) by 3% with respect to academic papers.
- Implemented progressive resizing in a Res-Net 34 CNN to develop an image classifier (~92% accuracy) from 3000+ images of eye retinas in fast/ai and Pytorch to speed up diabetic retinopathy detection in rural areas of India.
- Deployed models on several cloud functions. Developed a company UI which integrated these models to serve predictions in production at scale.

Python 3TensorflowPytorchGoogle ColabHTMLCSSJavascript ES6Fast.aiGoogle Cloud Functions

SK (SK엔카닷컴) Fall 2019
Software Engineer Intern - Deep Learning Seoul, South Korea

- Implemented MobileNetV2 CNN to develop an image classifier (93% accuracy) of 500+ different cars.
- Created an Automated Image Crawler to download and classify 10000+ images of different cars to train our Deep Learning model while detecting data imbalance.
- Deployed the classifier for production as a serverless API endpoint.

Python 3TensorflowKerasSeleniumGitScrumbanDockerNumpy StackAWS Lambda

Twitter Summer 2019
#EarlyBird2019 Participant (Software Development) San Francisco, California

- To make it easier to read all conversations around a Tweet, we proposed and developed the now deployed Retweets and comments feature.
- Explored monetization and implementations aspects as well as ethical implications.


The Church of Jesus Christ of Latter Day Saints September 2016 - August 2018
Finance Manager and Full-Time Missionary Salvador, Brazil

- Led a group of 10-15 co-volunteers spread throughout different regions. Conducted weekly training meetings to surpass our monthly goals by nearly 20% for three months in a row.
- Reduced operational time and costs by nearly 40% through developing and implementing computer programs that automated travel logistics plans, expenses record keeping and reconciliations for 200+ missionaries.

ExcelPython 3G SuitePortuguese Language