• Name matching project: Fuzzy matching millions of names using multiple stages (cosine similarity + MLP on (py)Spark)
• Name screening: Reducing false positives from name screening using Levenshtein distance, cos. similarity, Jaro-Winkler for feature extraction with LightGBM
• Anomaly detection (One-class SVM, Isolation Forest), address parsing (libpostal), database statistics monitor
• Snorkel labelling: Predicting label dependencies using robust PCA

I was involved in the process of building predictive models for investment risk in startups. The process involves:

• Retrieving data from different sources, including scrapping data from the web (using selenium with Beautiful Soup).
• Analysing the data, using Jupyter notebooks, pandas, matplotlib.
• Building Machine Learning models using Random Forest and SVM to fit our prediction problems.
• Feature selection/engineering.

I worked as Project Manager on the EU-Brazil project ATMOSPHERE.

I managed resources, following the status of the deliverables and delegating activities for three Brazilian Universities: UFMG, Unicamp and UFAM.

I was working on two different projects with the LaIC (Computational Intelligence Laboratory) research group:

1. Reducing the exponential size of solutions generated by the Geometric Semantic Genetic Programming (GSGP) framework. The main objective is to reduce the solutions in order to improve their interpretability and reduce memory and computational cost.

2. Analysing datasets used as benchmark for Genetic Programming (GP)-based methods under a Data Science perspective: We intend to gather datasets employed by the main publications in GP field in the last five years and analyse the viability of using GP to induce regression models.

I worked on the development of autonomous tugs, capable of towing the aircraft in the airport ground, from the runway to the gates—during arrivals—and vice versa—during departures.

Activities performed:

1. Surveyed current advances in autonomous vehicles and worked on optimization of ground routes in airports;

2. Worked in coordination with United Technologies;

3. Worked with Python (build a parser for XML airport maps) and Java/CPLEX (route optimization).

I worked with Geometric Semantic Genetic Programming (GSGP) on two main projects:

1. A study investigating aspects related to the semantic distribution of the functions employed by geometric semantic operators;

2. an investigation of the impact of different instance selection techniques on GSGP and its robustness to noisy data.

Comprehending:

• Management of a research team (three researchers);
• Development using Java (Genetic Programming framework), R (hypothesis tests and plotting) and shell script (text/data manipulation).