My goal is to detect fraud in money transfers within the payment market platform.- Develop a strategy to approve payments that, due to its high revenue, motivates us to take on more risks that we would otherwise reject the payment.- Tracking and maintenance of productive models in different flows.- End-to-end machine learning pipeline to detect fraud in transactions.- Creation of alarms and Business rules to detect deviations in the flow due to behavioral changes or attacks.- Detection of new behavior patterns for the creation of new features that improve the performance of the models.Main technologies: Datadog, SQL, Bigquery, Python, Excel, Google AutoML, Tensorflow.

- Demand Forecasting: End-to-end machine learning pipeline, which extracts the dataset, transforms all the features (ETL), creates new ones, and uses them to predict temporal series in a regression way. Uses of AutoML, MySQL, AWS S3 bucket, NiFi, and Redis.- Payment arrears forecast: Predict days of arrears in tax payments. In order to segment behaviors to take different actions for the client. Presentation of results with stakeholders.Main technologies: Python, Numpy, Pandas, scikit-learn.

Highly selective fellowship program (<2% acceptance rate) where top talent comes to accelerate their machine learning careers, surrounded by a community of AI professionals.- Shot Type Classification for Ads: Classify video scene shot types for ad performance measurement. The model returns two tags per frame: Shot Scale and Shot Movement. Deployed in AWS instances using Docker. Team Leader in this project.- Products review classification: NLP task classifying a movie review as if it is a positive or a negative review, using TF-IDF and Word2vec.- Image Classification for E-Commerce: Computer vision task classifying the car brand in an image using fine-tuned Convolutional Neural Networks and Detectron2 over almost 200 different brands. Deployed in AWS instances using Docker.- Deployed an ML model: Using Docker compose of 3 different microservices the web service with an API, redis to manage the queue, and the ML model to predict different classes using Resnet50 model.- Home Credit Risk Analysis: Using a Kaggle Dataset of Credit Risk to predict whether or not the user will repay their loan.- Salary Prediction Model: The goal was to predict salary levels based on historical data for sports players and predict the top 15 players to play in All-NBA using classification models. Collected and analyzed data via an API using Python and Pandas.

- User instruction manual: Update it with the new features developed for the other areas. Understand the technical content and explain it to the user. In three languages (Spanish, English, and Portuguese).- Validation and verification protocol: Performed different tests to verify the quality of the device. Create a new protocol that is more focused on the patient.- Risk management protocol: Assessed the importance of the different risks of the product in order to create more requirements decreasing the residual risk of having a safe health care device.- Usability test: Managed the usability evaluation by talking with doctors, nurses, and physiotherapists. Statistics were made to evaluate them.I generated a validation protocol that improved its execution time by 50% faster. Also design tests based on the end-user, evaluating the functionality of the equipment.

- Resolution of practical guides - Class planning

- Resolution of practical guides - Class planning