Architected and implemented an advanced RAG system utilizing ChromaDB for contextual embeddings storage, integrating OpenAI and Claude for enhanced semantic processing.Developed a hybrid search system combining semantic search with BM25 text search via Elasticsearch, implementing a Cross-encoder re-ranking model from Hugging Face, achieving a 98.1\% retrieval accuracy.Built a multilingual document processing pipeline using Hugging Face models for language detection and facebook/mbart-large-50-many-to-many-mmt for translation, enabling cross-lingual RAG capabilities.Created specialized AI agents for medical test interpretation (blood, urine, genetic) using the RAG system to provide contextualized health recommendations and insights to patients.

Developed Recommendation Models API: Spearheaded the development of an API hosting financial recommendation models. This API integrates the entire business logic and necessary data to provide clients with accurate portfolio movement recommendations.AWS Infrastructure Setup: Implemented and managed the entire cloud infrastructure using Terraform on AWS, including servers, databases, and S3 storage, ensuring robust and scalable solutions.Automated Data Updates: Utilized Django to develop APIs that enabled seamless communication with a dashboard built on Apache Superset. This setup facilitated automatic monthly updates of client recommendations by synchronizing with the database.Interactive Brokers Integration: Integrated Interactive Brokers to manage clients' portfolios individually, executing trades based on algorithmic recommendations at the beginning of each month, enhancing operational efficiency and client satisfaction.

360 Latam is a conglomerate of companies in Latin America and at this moment they have the objective to unite the biggest companies in Latin America in terms of real estate listings FincaRaíz and Infocasas. I led the integration and migration of data and apis to work from the Python backend with Django to the PHP backend with Laravel. This allowed the correct functioning of solutions that presented each business logic from code and also allowed these logics to complement each other and work in such a way that could maintain and boost the services provided by these two companies, for their customers.

As the lead Data Scientist for Autumn Lake, a pioneering nursing service provider operating across various states in the United States, I spearheaded the implementation of a comprehensive data analytics framework that significantly enhanced operational insights and decision-making capabilities. Utilizing my extensive expertise in data analysis and technology, I was instrumental in developing and deploying a suite of advanced analytics solutions that drove strategic growth and operational efficiency.Key Achievements:Strategic KPI Development: Formulated and integrated a robust set of Key Performance Indicators (KPIs) and metrics designed to meticulously monitor and evaluate profit margins and operational costs on a monthly basis. This strategic initiative enabled the executive team to identify trends, optimize resource allocation, and enhance overall financial performance.Advanced Analytics Dashboard: Led the design and development of a sophisticated analytics dashboard using Power BI, which provided real-time visibility into critical business metrics. The dashboard offered a granular breakdown of data by state, region, facility, unit, nurse type, and more, empowering stakeholders to make data-driven decisions with precision.Custom Web Portal: Spearheaded the creation of a dynamic website built on Django, serving as a centralized platform for real-time data interaction and analysis. This initiative provided company personnel with direct access to vital statistics, fostering an environment of informed decision-making and strategic planning.Predictive Modeling for Revenue Forecasting: Developed a cutting-edge revenue forecast model leveraging Scikit Learn, which accurately predicted monthly revenue streams based on historical data. By identifying key predictors such as patient numbers, nurse counts, job roles, and schedules, the model provided valuable insights that optimized workforce management and operational logistics.

Developed a web application using Django orchestrated with Docker Compose to integrate client database monitoring and usage. The objective was to enable clients to obtain results, insights, business question answers, dashboards, etc., by integrating GPT-4 with their data. An AWS EC2 instance was launched, where Docker was utilized to deploy Nginx services for traffic redirection and Python for the web application serviceI developed a web application to automate the creation, taking and summarizing of surveys for a client called HDO. The web application was created in Django, which connects to a Typeform API for taking the surveys; once the survey is filled in, the data is saved in a PostgreSQL database and then the results are displayed in a Power BI instance in order to see a simplified and aesthetic summary of the survey data. All these instances were created from Google Cloud using Terraform and their configuration was done with ansible.I implemented predictive analytics and inventory optimization solutions for a retail chain called OptiRetail using PySpark and Azure. I developed effective machine learning models, integrating MLOps practices for efficient deployment and monitoring. This project resulted in a remarkable improvement in inventory management and personalization of the customer experience, driving a significant increase in sales and operational efficiency

I led the deployment of a client's backend and frontend application, the deployment was performed following the Azure CI/CD process in which all the steps to deploy in the development, QA and production environments had to be instantiated in an app service for each environment respectively.

Finca Raíz as a client of Imagine Apps that characterizes its business model in real estate publications in a web application built in Django. I led the development of the implementation of the real estate business model by headquarters (branches), manipulating their endpoints and data models so that these customers could manage their products, the visualization of their statistics and control their publications, all through these sections of the real estate; achieving improved productivity and distribution of customer information.In a collaborative development, an endpoint was created for data consumption to display statistics and conclusions for the virtual office of real estate agents and builders. This allowed clients to have a concise summary of their business and an understanding of the market at the level of their listings.

Engaged in a technical role with a focus on learning and implementing DevOps methodologies, microservices architecture, Docker and Kubernetes integration, testing practices, and AWS cloud computing. Placed strong emphasis on enhancing soft skills and fostering a collaborative work culture, while actively exploring agile methodologies.During this internship, undertook a technical challenge involving the integration and deployment of applications using IBM App Connect Enterprise and IBM MQ with Docker. Successfully demonstrated skills in integrating applications and manipulating message queues, resulting in a positive evaluation and recognition as a semi-senior developer.I was selected to join a project with Bancolombia, specifically in one of their EVC (Continuous Value Delivery) teams. Assumed responsibility for developing comprehensive dashboards using power BI that showcased delivery statistics for the IBM Sterling team clients. Utilized a SCRUM methodology to efficiently monitor and improve file delivery, both on-premise and in the cloud.

I successfully led the development of Mileto, a web application for short-term energy demand forecasting and fault detection in photovoltaic systems, as part of my degree project. Utilizing Django, I built both the backend and frontend components, enabling users to train, test, and validate machine learning models without coding requirements. The application utilized a PostgreSQL database for structured data storage and integrated Google Drive instance through Google Cloud for automated model saving. By leveraging Docker-compose, I ensured seamless deployment of the web service, database, and Nginx server for production on Azure. Developed machine learning models demonstrated satisfactory performance in predicting energy demand and stock market trends. We achieved a Mean Absolute Percentage Error (MAPE) close to 5%, a strong outcome given the inherent complexity and volatility of both fields. Additionally, we were able to implement these complex models in a user-friendly web application, making them accessible to users with no technical knowledge about machine learning.I developed a web application and Restful API that enables users to upload inventory templates in .csv format to Azure Database for PostgreSQL. The system efficiently stores files in Azure BlobStorage and updates the corresponding data in the database model. Leveraging a comprehensive folder system, the application ensures easy identification of processed files by client, enhancing data management.