Leandro Lima

Leandro Lima

Leandro Lima

Machine Learning Engineer

Welcome!

My name is Leandro and I’m a Machine Learning Engineer with a passion for leveraging data to solve complex problems.

 

About Me

With a background in Computer Science, I have started working with Machine Learning in 2007, during my Masters degree. After that, I used Machine Learning techniques to identify patterns in patients with breast and prostate cancer. During my PhD, I integrated and analyzed DNA and RNA data of patients with ADHD to identify patterns in the genes and biological functions underlying the disease.

In 2016, I started working at Gladstone Instituteswhere I used Machine Learning to predict which and how patients developed ALS and Parkinson’s disease. As a Data Scientist working with human DNA and RNA data, I built an end-to-end pipeline to:

  1. extract, clean, and transform features
  2. train, test, and validate different ML models
  3. use SHAP values to check feature importance
  4. compare the different models (features, samples, hyperparameters and performance metrics) using Weights & Biases and Neptune.ai (to be visualized by biologists)

 

In March 2022, I joined Meta as a Machine Learning Engineer. In my first team, I worked with video recommendations for new users on Facebook. I built and maintained data pipelines (SQL + Python), and set up and monitored experiments (A/B tests) for millions of users. I launched a few models that increased the number of sessions and watch time without regressing other safety and business metrics (e.g. video recommendations based on implicit interests, and video recommendations based on engagement of closest friends).

In my second team, I worked on a project to predict the popularity/virality (number of views) of most of the content (videos, photos, posts, stories, etc.) on Facebook and Instagram. I was responsible for:
  1. Creating/removing feature logging (PHP)
  2. Adding new features to the model (Python)
  3. Training/testing the model offline
  4. Testing the model online (for a small, but increasing, percentage of clients/API requests)
  5. Checking several metrics to ensure the model could be deployed to all clients, and
  6. Deploying the model to all clients if the metrics look good.

 

Skills

•   Programming in Python (including Django and Flask), R, Bash, Javascript, C, PHP, Java, Perl

•   Machine Learning (Scikit-Learn, PyTorch, Tensorflow)

•   Statistical analysis, complex networks, Natural Language Processing, Recommender systems

•   Linux/Unix systems and high-performance parallel computing environments

 

•   Relational database management systems: MySQL, Postgres, SQLite

 

Teaching (Data Science and Computer Science)

I was a Computer Science lecturer for 4.5 years. I taught the following subjects at a university (FMU) in Brazil:

  • Graph Algorithms
  • Design and Analysis of Algorithms
  • Data Structures
  • Artificial Intelligence
  • C Programming
  • Game Development with Python
  • Software Testing
  • Object-Oriented Programming with Java

 

Independently, I’m working on a project called “Videos to Books“, to leverage LLMs to create books based on a video playlist.

 

 

Areas of Interest

While I have experience with recommender systems and general ML systems, I’m also interested in computer vision and large language models.