To develop 7 Recommendation Systems using Content-Based, Collaborative Filtering and Reciprocals algorithms using multi-processing. To develop 4 Image Classification algorithms using GCP, AWS and DeepDetect platforms for detection of sexual images, blur images, images without faces and famous images. To develop 3 Texts Classification algorithms using GCP, AWS and DeepDetect platforms for detection of sexual texts and texts with personal information like phone numbers and social networks. Data Visualization using different metrics, like: Retention Rate (Daily, Weekly and Monthly) and Conversion Rate (Daily, Weekly and Monthly) using Google Data Studioand Matplotlib (Python library). Precision and recall analysis using some techniques of "Retention + Engagement Deep Dive Reforge program" using correlations, logistic and linear regressions.

Re-train of personal loan prediction models using XGBoost, Random Forest and Logistic Regression, improving precision by 2% (from 96% to 98%) and recall by 5% (from 81% to 86%) on average using approximately 8,000 loans and 32 features. Creation of a new template pipeline in order to deploy ML models improving the deploy time of a new model by 75% (from 1 month to 1 week).

Creation of a datamart for Data Analysis using Google Cloud Platform tools such as Big Query, Google App Scripts, Google Storage and databases such as MongoDB. Implementation of a new process to deploy ML models using GCP tools such as App Engine, Google Storage, Google Datastore, Google Dataflow, Google ML, Cloud Datalab and Google Vision. Data Visualization using different metrics, like: Retention Rate (Daily, Weekly and Monthly) and Conversion Rate (Daily, Weekly and Monthly) using the Google Data Studio tool and Matplotlib (Python library). Precision and recall analysis using some techniques of "Retention + Engagement Deep Dive Reforge program" using correlations, logistic and linear regressions, improving the main KPIs like retention by 5% (from 38% to 43%) and conversion by 7% (from 0.1% to 0.7%). Deploy of sexual, group and blur images classifiers algorithms using Google Cloud Machine Learning Engine library, inception networks and transfer learning over 10000 images. This improved the approval time of profile images from 4 hours to 5 seconds and their precision by 87% on average for the 3 models. Development of a Recommendation Algorithm using Reciprocals Methods for categorical variables and a Deep Learning approach called Paragraph Vectors (DOC2VEC) for free text variables. This new process improved the retention by 4%.