Hi, my name is Tiago dos Santos and I’m a curious mind with around 6 years of working experience. I’ve finished my Master’s in Computer Science and now, after a career break, I am looking for a challenging environment where I can discuss ideas, learn state-of-the-art techniques, and where I can make the difference in the Data Science field – being it modelling a problem and/or engineering and implementing the algorithms in a (near) production-ready fashion.
The Master’s I’ve took gave me the opportunity to develop projects and acquire further knowledge in areas of study such as distributed systems, parallel/concurrent processing, data processing / warehousing, data visualisation and machine learning. My Master’s final grade is 17 out of 20.
Since 2016 I’ve been focused on Data Science, which is a subject for which I find myself pretty curious about and with a variety of topics that make me eager to learn. The first professional project where I’ve been involved was my Master’s Thesis Dissertation, which was scored as 20 out of 20 (it was the sixth 20/20 grade since 1976 in the course).
The goal of my Master’s Thesis was to develop a model capable of detecting Stator Winding Short-Circuit faults in Induction Motors. To do so, I had to study electric circuits and electric machines, particularly three-phase induction motors and their working principle. After the data understanding phase and a state-of-the-art study, an assessment about the predictive power of extracted features was performed using a SVM. Given the best set of features, the most promising models for this context were determined, where after a parameter and hyper-parameter tuning the Gradient Boosting Machine and Random Forest revealed both the best performance. Due to its inherently parallelizable nature the Random Forest was chosen. The raw dataset had 29440 instances with 6 features, but due to modeling options the processed dataset had 1472 instances with 25 extracted features. All the analysis was made in R and the documentation made with Bookdown.
To automatize the process of data fetching and data understanding, I developed an ETL service (node.js) as well as a dashboard for data exploration and filtering (R, Plotly, Shiny). Since my Master’s Thesis Dissertation was developed within an Industry Partnership with Altran Portugal, I was also responsible to present/explain my research on a regular basis to the company’s Managers not only to keep the staff updated about the on-going R&D projects, but also to guarantee European R&D Fundings.
Outside of my Master Thesis’ context but in the context of the Altran Portugal’s Project, I’ve also develop a non-stationary learning model based on the Thesis’s model, in which the dataset had 2000 instances with 50 extracted features.
My Dissertation culminated into two papers:
- “Stator Winding Short-Circuit Fault Diagnosis in Induction Motors using Random Forest”, which was presented at IEMDC 2017, Miami.
- “Severity Estimation of Stator Winding Short-Circuit Faults using Cubist”, which was presented in EPIA 2017, Porto.
During my thesis, I’ve started taking the Machine Learning Coursera Specialization. From this specialization, I’ve already conclude the following courses:
- Machine Learning Foundations: A Case Study Approach
- Machine Learning: Classification
- Machine Learning: Regression
From this specialization I’ve only one course left to finish, which is Machine Learning: Clustering & Retrieval.
During my Master’s, I had the opportunity to negotiate, close, lead and develop a project as a freelancer where Fluxphera company needed an online platform to provide a multiplayer gamification process and to manage the users and their data. One of the requisites of this platform was the scalability, since the workload during the day was variable (docker, docker-compose, node.js). Due to the lack of IT collaborators on the company, the delivered solution was based on the AWS EC2 computing instances: an instance where the Load Balancer Service was running, and (by default one) instance where the web app service was running. When the workload of the web app service instances was higher than a given threshold, another web app instance could be launched.
I’m also an active listener of the Data Skeptic Podcast (I’m a Patron of the project) and I’m an enthusiast of the youtube channels 3Blue1Brown, MITCSAILS, and TwoMinutePapers.
I like to stay updated, and to do so my mainly reading routines consist on Hacker News (Y Combinator), DataTau and Reddit.