Logo

George Saad

I'm a Master's student at the University of Toronto, doing research in Large Langue Models (LLMs) and Recommender Systems (RecSys) with Prof. Scott Sanner. I have extensive experience in AI & software through many projects and internships.

Experience

U of T Data-Driven Decision Making Lab (D3M)

Research Assistant

May 2023 - Present
  • Working on integrating information retrieval and LLMs (such as GPT-4) with conversational recommender systems to improve the state-of-the-art in recommendation at Scott Sanner’s Data Driven Decision Making (D3M) Lab

University of Toronto

Teaching Assistant

September 2021 - Present
  • Teaching Assistant for: Introduction to Programming (ESC180) in Python, Algorithms & Data Structures (ESC190) in C

Vector Institute

Applied Machine Learning Intern

January 2023 - September 2023
  • Implemented metrics to evaluate the performance of causal inference estimators, including TARNet and Dragonnet, on synthetic datasets and observational data with no ground truth counterfactuals
  • Led lectures & hands-on sessions on causal estimation & recommender systems for 200+ technical participants, including data scientists and executives from 40+ teams across major sponsors such as RBC, Deloitte, Shopify & Air Canada

Meta

Software Engineer Intern

May 2022 - July 2022
  • Designed Thrift service to add tracking to over 260 million external Instagram story and profile links per day
  • Implemented client-side app detection feature and ran controlled experiment on 4% of users to determine impact, which resulted in a 120% increase in ad impressions for Instagram users without linked Facebook accounts
  • Performed major backend code improvement in Hack (PHP) and Python Django for the Instagram Ads endpoints

aUToronto

Mapping Software Team Member

September 2021 - April 2022
  • 1st Place Winner of 2022 SAE Autodrive Challenge II
  • Maintained testing track map in OpenStreetMap using JOSM and implemented Python methods with PyOSM library

Loblaw Digital

Software Development Co-op

May 2021 - August 2021
  • Set up latency log tracking in various functions in the Spring Boot Java backend and created GCP metrics and dashboards to extract and display the data, with 200k+ data points collected per hour
  • Led research into content targeting engine and presented to the 40+ member PC Optimum team to enable personalization of marketing tiles for consumers
  • Created various internal self-serve and admin endpoints to reduce development, testing, and debugging time by over 60% for various parts of the PC Optimum product

Content Turbine

Software Engineer (Freelance)

December 2020 - September 2021
  • Built NoSQL datastore and caching modules for the Akka Play! and Vert.x frameworks in Java using Singleton and Dependency Injection (DI) design patterns, and reactive programming, with Guice and JUnit unit testing
  • Contributed to the JHipster open-source project (over 18,000 stars) by implementing a major upgrade for a Spring Data SDK integration, including Full Text Search (FTS)
  • Wrote ASCIIDOC documentation, and created a demo app and an introductory blog post

OrangeTopi (now FoodTrain)

Lead Software Engineer Intern

May 2020 - December 2020
  • Led a team of 8 – 10 developers in developing a React Native mobile app MVP in under 6 months by providing guidance and developing components, including app notifications and authentication with the React Context API
  • Reduced deployment time by over 90% by implementing a CI/CD pipeline for automatic deployment
  • Developed multiple backend functions in Node.js and Express.js, including all payment and email handling

Delovery

Software Engineer Intern

March 2020 - June 2020
  • Reduced market research time by over 80% by building a Python web scraper using Beautiful Soup to scrape and graph 900+ data points using Matplotlib
  • Designed and developed the cart user flow and other major frontend components in React.js with Material-UI
  • Created an API using Node.js and Express.js to handle all payment processing with PayPal Braintree Payments

Projects

DeepDixit

April 2022
  • Built an image caption guessing game using SIREN + BigGAN based networks to create an image generator with a CLIP-based scoring function to fine tune the model based on text prompts

micromalaria

December 2021
  • Fine-tuned a ResNet classifier to identify whether cell images were infected with malaria for an African health lab
  • Achieved a 99% decrease in false positive rate from 19.4% to 0.16% and a 17.5% increase in accuracy from 80% to 94%

crowd.data

January 2021
  • Top 4 out of 30 projects submitted for the Dropbase API prize
  • Developed Node.js and Express.js RESTful backend and React.js frontend components with Chakra UI
  • Created 4 data models with CockroachDB (PostgreSQL) using the Sequelize ORM, and Dropbase API integrations

GPT-3 for Finance

July 2020
  • Built a RESTful Python Flask server to create and fill a balance sheet based on natural statements, using the OpenAI GPT-3 NLP API and the Google Sheets API
  • Received 100,000 views, featured on InfoQ and on Y Combinator Hacker News front page

HyperBot

January 2020
  • Won 1st out of 70 teams by building the best healthcare chatbot (Hypercare API prize)
  • Employed Google Cloud App Engine to host a RESTful Python Flask backend to receive POST requests from webhooks and Google Firebase to store and update a Firestore database using JSON files
  • Implemented Hypercare API integration to receive/send messages, schedule appointments, and find doctors
  • Used a Python ELMo module to preform NLP on user input and map it to a symptom to produce a diagnosis