Mi Aguila is a transport, logistics, and eCommerce technology company focused on helping traditional businesses in Latin America thrive in a digital world. My main role here consists in developing the ETL process through AWS services that can be easy consumed by the BI team and customers. Also, I supported the rest of the team improving metrics and taking decisions. Main Technologies & Tools involved: AWS services (S3, Lambda, Glue, RedShift, Athena, Clowdwatch, Step functions, EventBridge, Kinesis, Ec2, Jira, Confluence, Metabase, Amplitude, among others). Team size: 1 Data Analyst + 1 Data Scientist + 1 Data Engineer + 1 TL

Credin Transfer Gateway is a module that re-directs petitions of transfers between bank accounts, especially the ones that are not within the same bank when third-party services need to be consumed to get the money transfer done. In other words, it directs transfers between BICA accounts (source and target accounts) and to accounts in other banks (source account in Bica and target account in another bank). The module was built to expose a REST API that supports individual transactions (one-to-one) and transactions in bulk quantities (a source account and any number of target accounts). The source account is always an internal account and the target could be either internal or external. We also had to design and build a component to allow legacy applications built in VB6 to connect to the REST API. My task in the project consisted of defining and developing different modules to be able to perform the process above mentioned. Team size: 5 developers + 1 Project Manager

FoF is an e-commerce app that recollects data from multiple vendors in real-time and performs weighted comparisons between products according to user preferences. It has a microservices architecture consisting of the crawling engine, the comparison engine, the database, the frontend, and the backend/API. It also features requests for proposals, where users can define requirements and vendors can create specific offers from the vendor dashboard. The project consisted of creating all the microservices and integrating them, crawling the vendors from the webpage or using the APIs, and creating automatic tests that detected outlier data. All the user and security management was created using the latest OWASP recommendations as of 2019. The frontend is responsive and can be used from mobile and web browsers, and features a live chat using WebSockets so vendors can contact individual users to follow up on sales or to provide clarifications.

This project is a port of a fully-fledged, multi-tenant ERP that includes different modules,sales, human resources, billing and more. This product is fully adapted to the customer's needs. My work here was to fix some of the rough edges of the old system and gave it to a new stack of technologies to simplify its maintenance and speed up the development process, improving, at the same time, key quality attributes like reliability, usability, and security.One of the main challenges of the project was the complete redesign of the architecture of the old system to be able to give the clients the top quality they deserve. The new architecture is based on a three-tier client-server style pattern with tenant-separated databases.The main areas of my work are: - Front End & Back End. - Integration between stack layers. - DevOpsTeam size: 4 Developers + 1 Project Manager

The aim of this project was to analyze the use of two different services of Amazon WebService called Comprehend and Sagemaker. The client has its own NLP algorithm based on ApacheOpenNLP which accomplished the detection of some custom Named Entity recognization (NER). To improve the performance of NER detection and also move the initial project to a cloud-based solution, the decision was to start working on AWS as a maint cloud solution. Two projects were made using, first AWS Sagemaker and the standard NLP models, and second, AWS Comprehendcustomizing the detection of NERs regarding the client's specifications. Once the model for NERs detection was deployed, AWS Comprehend was configured to be used as a Document Text analyzer, providing the corresponded API Endpoint to use the service. This Comprehend Endpoint was wrapped on a custom API service made with Flask (python).

This project aims to achieve an Image Recognition System able to distinguish between the different items with which the company works, and count the number of items in case it was possible and finally determine the expiration time of these products base on visual impairments. Starting with the module able to distinguish between the different products We decide to use a pre-trained neural Network taken from the Keras library, based on the Net has been used on similar works and brings the possibility of using on mobile phones. Then through testing the Net, I determined the appropriate measurement of the images to be able to work them. I use Adam optimizer and cross-validation from the Tensorflow library. The results obtained were optimal: Precisions above 99%. This generated a pleasant surprise and great acceptance by the client. In the second stage, we had to obtain a functional model able to segment the image into multiple sites to count them.