The first visualization consisted of a static infographic, comparing the usage of two docking stations during January 2020. The infographic was designed to show the performance differences of two stations with opposite characteristics.

The two compared stations where Station 31, at sea level and Station 327 at the mountain. At mountain level stations there’s high demand of bikes, because all the routes are downhill. However, at sea level many people arrive with bike, but there’s not a high demand for bikes. This causes during peak hours a deficit of bikes in mountain stations and a deficit of free docking space at sea level stations.




The second visualization is a set of interactive plots, showing the performance of all Barcelona stations during January 2020. The aim of this more complex visualization was on one hand to show an “overview” of the whole Bicing service and on the other hand to allow the user access specific information under request. With the use of Altair interactive capabilities, more information can be shown in an intuitive way, making it easier for the user to understand the data and find meaningful relations.



This project was developed in the context of VI -GCED class, as a way to implement both theoretical and practical knowledge learnt during the course. All the visualizations were designed with a strong focus on technical quality principles such as Schneiderman’s mantra and the laws of perception.

Both visualizations are implemented in Collab ipynb notebooks. In the notebooks all the design process, decisions and final result are detailed (in Catalan). It can all be found in this repository.

This project was developed with the goal to study in which way kernel functions could better prediction results, and discover ways to fine-tune custom functions in order to to maximize the performance.

This work was done in the context of AA2 UPC class final project, oriented to kernel methods. Most of the work was done with Python’s sklearn and R for preprocessing.

The final report with conclusions and interactive notebooks with used code can be found at the github repository.

The objective of this project was the creation of a distributed ETL data flow for an aviation maintenance business case. The original dataset is a set of maintenance and flight data from a real airline (AIMS and AMOS extracted data), and the goal was cleaning the data and extracting the necessary KPI’s in order to train a model to predict whether maintenance was required for a specific airplane.

All code was done having distribution in mind, with Python and Spark API, pyspark. Prediction was done with mllib.

The final result is a script that can be executed with a terminal, that is capable of training a model with data extracted online and predict on-demand with user input data and different configurable characteristics

The code developed and extra information on how to use the tool can be found at this github repository.

The first project was oriented on learning to work in HPC (High Performance Computing) distributed environments. We where given access to the MareNostrum supercomputer with the aim to create and execute scripts to run the NAS Parallel Benchmarks with different configurations. The scripts and final report can we found here.

The second project objective was to learn to work with cloud providers, taking advantage of its distributed capabilities. The goal was to extract twitter analytic information with the twitter API and processing it using Spark, distributed in several AWS EC2 machines. The resulting report and code can be found here.

The application is used via direct chat with the bot. The network of stations is modeled as a graph, that is created on demand based on user input specifications. The bot is capable of calculating fastest routes between origin and destination points, printing a map with the path. It’s also capable of calculating the redistribution of bikes, given the number of required bikes per station returns the cost of doing a transaction to solve it.

The bot was developed with Python, the NetworkX library was used to model the graph and StaticMap to plot the maps.

TheRealBicingBot was developed as a final project for AA2-GCED class. The complete code, with more detailed documentation and testing functions can be found at this github repository (in Catalan).

The user can input polygons as a set of points, and calculate metrics such as area or perimeter. Multiple polygons can be inputed, so it’s possible to generete intersections and bounding boxes.

All results can be saved in a .txt file and loaded again. The program also supports plotting inputed polygons and results in a .png image, with the pngwritter library. All code is created with optimization in mind, using several complex algorithms to keep computational costs low.

This project was developed for AA2-GCED class, as a way of testing C++ and algorithm knowledge learnt during the course. The complete code and extensive documentation can be found at this github repository.