# Javier Saez Gallego

## Data Scientist and Developer

Javier is a passionate data scientist who builds data-driven software products and helps organizations transform numbers into optimal decisions. With a multidisciplinary background, he has analyzed highly dimensional datasets and built machine learning models—starting from the research stage to the implementation phase. Javier is careful, very organized, meticulous about planning always focusing on delivering the right solution.

### Portfolio

### Experience

Data Science - 9 yearsMachine Learning - 9 yearsOperations Research - 6 yearsLinux - 5 yearsR - 5 yearsPython - 4 yearsDocker - 2 yearsAmazon EC2 - 2 years### Availability

### Preferred Environment

Linux, Docker, Git, Python

### The most amazing...

...thing I've done was to develop an application used by sales engineers for the warranty calculation of wind turbines.

### Work Experience

#### Data Scientist

##### Reforestum

- Developed deep learning models that monitor forestry areas using satellite images. The final goal is to monitor forest conditions and to calculate carbon stocks in real-time.
- Implemented a back end that performs ETL, machine learning modeling, and serves the results via an API.
- Visualized GIS results via Mapbox.
- Trained and deployed machine learning models in AWS.

#### Data Scientist Consultant

##### Minsait

- Identified potential business cases where machine learning models could bring value to the clients.
- Participated in the research, development, and implementation of predictive maintenance models of ATMs for one of the biggest banks in Europe.
- Developed dashboards and interactive graphs.

#### Data Scientist

##### TecDeSoft

- Created business value with available data from existing clients, focusing on data-driven action.
- Implemented predictive maintenance of factory areas and optimal scheduling of hydroelectrical power plants.
- Taught my colleagues the principles of data science and machine learning.

#### Data Scientist

##### Siemens Gamesa

- Developed a framework consisting of a database, an app, and a dashboard in order to let users predict the performance of wind turbines.
- Analyzed data from measurements and detected possible misalignments with the expected behavior.
- Extracted meaningful information from multi-dimensional datasets.
- Communicated the results and statistical terms to electrical engineers and sales officers in a clean and concise manner.
- Read, studied, and kept up to date with current ISO standards.
- Quantified the risk of different warranty strategies.

#### PhD Candidate

##### Technical University of Denmark

- Developed models for decision making under uncertainty, based on stochastic optimization techniques.
- Built-up a new type of forecasting modeling framework that was based on inverse optimization techniques and machine learning principles.
- Extensively used R for data processing and GAMS-CPLEX for building optimization models.
- Used a cloud computing framework for parallelizing the calculations.
- Presented the research topics and results in international conferences in Lisbon, Glasgow, and Philadelphia.
- Published four articles in well-ranked journals.
- Worked as a visiting scholar at the University of California for four months.

### Experience

#### Squirrel Problem

http://jsaezgallego.com/GlobCover_maps_squirrel/The answer is: obviously not. But, another question arises, and this one is not so easy to answer: If a squirrel had to go from the north of Spain to the south, touching the ground as little as possible: Which way would it follow?

The answer is not trivial, and it is answered here!

The project consisted of the following steps:

1. Download data from the internet. The data consists of GIS information to determine whether a piece of land is a forest, a river, a man-made construction, and so on.

2. Create a path matrix. The squirrel takes one step at a time. In terms of a raster map, this means we can reach each pixel only from adjacent pixels. Think of it as a huge spare matrix.

3. Optimize the path. Calculate the shortest path from the top of Spain to the bottom. The cost of each segment is the pixel value so forest costs nothing to move across and anything related to water is impossible to cross.

4. Visualize the solution as an interactive map.

#### Notification Bundler

https://github.com/jsga/bundle_notifications### Skills

#### Languages

Java, Python, R, JavaScript, SQL, GAMS

#### Frameworks

Spring, Flask, RStudio Shiny

#### Libraries/APIs

Node.js, Pandas, Plotly.js, Keras, Ggplot2, TensorFlow, React

#### Paradigms

DevOps, Data Science

#### Other

Software Development, Machine Learning, Data Analysis, Dashboard Design, Operations Research, Mathematical Modeling, Computer Vision, Algorithm Development, Optimization, Dash

#### Tools

Git, MATLAB, Plotly, CPLEX, Dplyr

#### Platforms

Linux, Docker, Amazon Web Services (AWS), Jupyter Notebook, Mapbox, Amazon EC2

#### Storage

Amazon S3 (AWS S3)

### Education

#### Master's Degree in Mathematical Modeling

Technical University of Denmark - Copenhagen, Denmark

#### Master's Degree in Statistics and Operations Research

University of Valladolid - Valladolid, Spain

#### Bachelor's Degree in Statistics and Operations Research

University of Valladolid - Valladolid, Spain

### Certifications

#### Deep Learning

Deeplearning.ai via Coursera