Skip to content

Saullo Oliveira

I am a data scientist, currently finishing my PhD at the Laboratory of Bioinformatics and  Bioinspired Computing,  in the School of Electrical and Computer Engineering, on University of Campinas (UNICAMP); advised by Prof. Dr. Fernando J. Von Zuben.

My current research focuses on Structure Estimation for Multi Task Learning, where we design algorithms able to learn several tasks (for example classification or regression tasks) simultaneously, while estimating an interpretable relational structure among those tasks.

My research interests include:

  • Multi task learning
  • Sparse models
  • Convex optimization



Programming Languages

C / C++


  • scikit-learn
  • pandas
  • Numpy
  • TensorFlow
  • Keras
  • matplotlib
  • seaborn

Machine Learning

All general stuff like Least Squares and variants (Ridge Regression), Lasso and variants (Group Lasso, Fused Lasso, etc), Logistic Regression, Support Vector Machines (SVM), k-NN, Decision Trees, Neural Networks, etc.

On clustering I have experience with K-means, Spectral Clustering, Agglomerative Clustering, DBSCAN, Mixture of Models, etc.

During my master’s degree and beginning of my PhD I’ve worked with Biclustering, Formal Concept Analysis (FCA), and Frequent Pattern Mining. 

Some tools include InClose (discrete), RInClose (real), Triclustering, MicroCluster, etc.

This is my area of research nowadays. We can play indefinitely here =).

I’ve studied and applied several ideas in this area, like:
  • MLP
  • RBF
  • ELM
  • Echo State Networks
  • Recurrent Networks
  • Constructive Networks
  • Kohonen Maps
  • Auto-Encoder
  • Convolutional Networks
  • U-net
  • etc

This is a topic that I’m really into recently, having some experience reducing non-smooth, multi-convex problems into smaller convex sub-problems.

Some methods that I am familiar with include:

  • Gradient Descent (and most common variants)
  • Accelerated methods: ISTA, FISTA
  • ADMM
  • Quasi-Newton methods: BFGS, L-BFGS
  • Others: Levenberg-Marquardt, Cache Factorizations, Debugging and Profiling…

The sky is the limit. Seriously, several things to write, do later…

I have excellent skills in all stages of a ML Pipeline, including feature engineering, data wrangling, parameter estimations, integration of data models into data products for high volume data.

Get in touch using the Contact page =).


Profiling and Code Optimization
Test Driven Development
Agile Development