Panamá, Panama
• Embedding Models for RAG: Familiar with fine-tuning transformer-based models using frameworks like llamaIndex and LangChain. I’ve worked with evaluation metrics such as Mean Reciprocal Rank (MRR) and Recall@K to track retrieval improvements, ensuring better accuracy and reduced irrelevant responses. • Explored synthetic data generation techniques to address data scarcity issues, improving model generalization in low-resource domains. • Developed and deployed AI-powered APIs to expose machine learning models to internal and external stakeholders, ensuring code reusability and modularity. Optimized API response times to under 100ms, improving system efficiency by 30%. • Collaborated with stakeholders to gather and refine technical requirements, translating business needs into production-ready applications. This application required a frontend built with Plotly Dash to enable user interaction with the algorithm and a backend built in Go or Python for efficient processing and model execution. Implemented API rate limiting to ensure scalability. Optimized backend performance, reducing API response time by 40%, and integrated logging and monitoring tools for real-time system insights. • Implemented MLOps best practices, reducing model deployment time by 60% through automated CI/CD pipelines (GitHub Actions). Integrated monitoring solutions using Grafana and Pyroscope for real-time model performance tracking. • Productionized and scaled data science models, integrating them into robust inference pipelines. Implemented auto-scaling mechanisms that improved model availability by 99.9% uptime while reducing compute costs by 20%. • Authored comprehensive technical documentation for model maintenance, API usage, and deployment workflows, ensuring seamless handover to future developers and reducing onboarding time by 50%.
• Designed and developed a high-performance data ingestion solution, enabling the simultaneous loading of large files into the Segment Platform. Built using Go and the Segment API, this reduced processing time from 12 hours to just 10 minutes, significantly improving data availability. • Developed an automated ETL pipeline, streamlining data collection and processing across 10+ countries. Built using BigQuery, Google Kubernetes Engine (GKE), and Python, this reduced a 1-month manual process to mere minutes, enhancing operational efficiency. • Collaborated with global business teams across Brazil, Canada, Poland, and Singapore to understand data needs and deliver tailored data products that drive decision-making. • Worked with the Global Data Engineering Team to enhance workflows, optimize data architectures, and improve the scalability and reliability of data pipelines for AI and analytics applications. • Built and deployed production-ready environments, enabling seamless deployment of machine learning models developed by the Data Science team, ensuring scalability, monitoring, and model retraining workflows. • Led cost optimization initiatives, reducing cloud storage costs by 20% through data partitioning, clustering, and efficient query optimization in BigQuery.
Identify new challenges posed in the implementation of Artificial Intelligence, BigData, data science, and other emerging technologies. Provide technical support with its capabilities in the successful execution of service contracts demanded by our clients.
• Develop an Object Recognition model capable of recognizing and classifying different business SKUs for a Share of Shelf study. This model reduced the time and effort implemented on manually counting business SKUs, by automating the detections and analysis of each shelf. • Design and build Proof of Concepts (POCs) using Object Recognition state-of-the-art algorithms. Testing and documenting each algorithm to help identify the advantages and disadvantages of each of one, this drives to a more meticulous and secure decision on which algorithm adapts better with the study case selected. • Design and build service application architecture on Google Cloud Platform, capable of hosting and Object Recognition model and receive requests through API end point.
• Design and develop an IA-driven Call Center service, using IBM Cloud environment and Watson assistant chatbot. This solution improved the attention services of the company by decreasing the response time and taking more than 30 concurrent calls instead of one. • Build an algorithm capable of identifying a name by its phonetic representation. This algorithm solved the problem of a voice-based chatbot that focused on client search by name pronunciation. • Develop data pipelines that consolidate reports from multiple business areas in one dashboard. These pipelines reduced the effort and time spent by each business when preparing reports, by automating the data transformation and consolidation process of business metrics and results.