Angel Ruiz Reche

A deep learning tool created using Python and Tensorflow that can rebuild genomes and evaluate their expression without having a reference from that species. It works by stimulating the ARN splicing, a biological process that uses a language (similar to DNA) that has not yet been fully deciphered. Therefore, ReorientExpress can predict the results of splicing without the explicit need to deciphering the splicing code.

This highlights one of the biggest advantages of deep learning; it can simulate complex systems without having to simplify the process into simple rules. Rather, it can learn complex interactions that other machine learning models cannot.

A deep learning-based application in Python that helps software testers by selecting the best samples to test. It first creates a model that tries to replicate the program (the Oracle) using deep neural networks. Then, given a new dataset, it selects the samples that are more likely to find bugs in the software under testing.

A Python-based application that connects to an AWS Athena service, extracts hourly advertising data from Amazon ad campaigns, and forecasts the traffic and conversion data of the active campaigns.

Those forecasts are sent to a web app in which clients can see which campaigns are likely to go out of budget during the day and by how much.

They also get an estimate on the potentially missed traffic and conversion events and a suggested budget increase to avoid going out of budget. As a result, their campaigns are always on budget.

I created a conversation-based videogame using GML, a language based on C++. I designed and coded the structure, the graphical interface, and all the behind-the-scenes logic.

In this conversation-based videogame, the user communicates with an AI assistant using text inputs and can perform several quizzes, psychological tests games, and more. It uses Google Cloud services such as:
• Storage: To store user behavior data and gameplay event data.
• Functions: To allow communication between GCP and the videogame. It uses several endpoints for specific tasks.
• A text-to-speech API: In combination with Functions, it allows the AI assistance to speak.

The game explores topics such as transhumanism, hedonism, and individualism.

I created an ETL system in which data from different APIs and an on-premise PostgreSQL DB were dumped into AWS S3 using AWS's Lambda Functions.

Then the data was parsed, processed, cleaned, and then uploaded to AWS's Redshift. Data was also homogenized so the different data sources could be queried together. The pipeline was scheduled to run automatically at midnight ever day. The process was fully logged and fully developed in just 4 days.

Finally, the data was connected to an external dash-boarding solution (Metabase) where it could be visualized in real-time.