Guarulhos, São Paulo, Brazil
As a former astrophysicist with a background in Physics and a PhD in computational astrophysics, I have spent the past eight years making significant contributions to the field with over 10 published papers, focusing on Machine Learning and Data Analysis. In 2022, I transitioned into the corporate world, where I now apply my analytical and problem-solving skills in the development of innovative solutions for the Healthcare, Agriculture and Applied Sciences industries. Specializing in Machine Learning Engineering, Deep Tech Applications / Architecture, and the creation of APIs and integrated services, I leverage data analysis and data science techniques to push the boundaries of what's possible, continuously elevating my skills and expertise. My most recent works include Apps and APIs using Large Language Models (LLMs) like GPT-4, as well as Speech Recognition Models like Whisper, to generate business value in the most strategic and data-driven aspects of organizations. Main Achievements: - Development of ERP System (front-end, back-end and infrastructure management side) integrated with data intensive calculations - Development of Real-time Speech Recognition services as well as Language/Content Generation services, using useWhisper React hooks as well as Streaming OpenAI hooks - PhD in Computational Astrophysics - Published papers with codes in the domain of Data Science - Published softwares (open-source) using efficient languages (Rust, Scala) for heavy numerical calculations - Full-stack App development specialization (from React + Next on Front-end to APIs and ORMs integrated with databases in the back-end) Core Skills: - Efficient Languages Software Development - Analytical and Mathematical Background from my Academic Experiences - Connect business needs with software development using Microservices (SOA)
At LumaHealth, I contributed as a Staff AI Engineer on Navigator, an AI-powered app designed for major hospitals and clinics across the U.S. The platform enabled seamless patient interactions through SMS and Voice APIs, covering critical use cases such as: - Appointment scheduling, rescheduling, confirmation, and cancellation - Automated handling of clinic-specific FAQs - Customization for each clinic’s working hours and appointment management policies I architected and developed the LLM-powered back-end solution, leveraging advanced multi-agent systems and orchestration patterns. My contributions included: - Designing AI workflows with LangChain, LangGraph, and custom multi-agent architectures (LLMCompiler, ReAct, GraphRAG, Supervisor–Worker) - Implementing Langfuse for LLM observability, prompt management, and evaluations in appointment management workflows - Integrating MongoDB, FastAPI, and Redis for scalable, reliable back-end operations (including caching and vector database functionality) - Ensuring HIPAA compliance and secure handling of patient information across all workflows The solution demonstrated how AI can safely and efficiently streamline healthcare operations, providing hospitals and clinics with a customizable, intelligent assistant to improve patient engagement and reduce administrative overhead. Keywords: LLMs | Generative AI | LangChain | LangGraph | Langfuse | Multi-Agent Systems | LLMCompiler | ReAct Agents | GraphRAG | Supervisor-Worker | FastAPI | MongoDB | Redis | Vector Database | Healthcare AI | HIPAA Compliance | Prompt Management | LLMOps
During my experience as a Full Stack Developer in the Benson Hill Data Analytics Team, I was involved in the development of an internal ERP system related to the Design of User Interfaces and APIs, along with managing the backend database to support the company's internal management services. The technology stack employed includes: - React and Next.js for front-end development - AWS CloudFront and AWS Route 53 - Python FastAPI and Lambda for backend integration - RDBMS, more precisely, based on PostgreSQL and hosted in AWS RDS - Flyway as DB Migration and Versioning Tool (DevOps for Databases) - Infrastructure as Code with Terraform - Automation and Continuous Integration/Continuous Deployment (CI/CD) using Bitbucket Pipelines, AWS CodeBuild, and AWS CodePipeline Main Result: Deployed a Stack for the Application (ERP System) with high availability and low-cost using AWS. the solution allowed the company to migrate from a legacy system to a newer, more efficient system oriented towards the specific business needs. Keywords: AWS | RDS | PostgreSQL | ECS and Fargate | Lambda Functions | FastAPI | React with Typescript | Next.js | Terraform | Bitbucket Pipelines
While working in the Benson Hill Agriculture and Biotechnology Data Analytics team as a Data Science Engineer, I have contributed to the projects of data and analytics pipelines by adding new Business Values and Features to the ingestion and data engineering pipeline. The pipeline was responsible for capturing raw data from sensors and crop fields and performing a series of transformations to predict future outcomes of existing crops. The pipeline involved using the phenotype of several plants to make predictions about market viability using Bayesian Statistics and Data Analysis. To achieve this goal, I used the following Cloud Resources: - Amazon SageMaker (pipelines for the data transformation and intensive parallel calculations due to the high number of crops and their individual states and characteristics) - AWS Glue (Crawlers and ETL Jobs to make certain types of data transformation, faster) - AWS Lambda functions and Step Functions (to orchestrate integration steps that were required in the middle of the pipeline) - Amazon CloudWatch and EventBridge (for orchestration and ESB) - Amazon Athena (for query execution and Exploratory Data Analysis) Main Results: At the end of this project, the new tools developed enabled BH to keep track of one of the core businesses of the company, as well as simulate scenarios of future field activities, impacting at the strategic planning level of the company. Keywords: AWS Sagemaker | Batch Jobs | DAGs | AWS Glue | Lambda Functions | Step Function | EC2 | Finite State Machines
At Siemens Healthineers, I work as an Application Enabler responsible for delivering scalable, resilient and user-centered applications using technologies based on Generative AI with modern architectures like Retrieval-Augmented Generation, Document Parsing and Multi-Modal interaction (Speech-to-Text, Image-to-Text, Text-to-Speech). The main objectives of the applications I develop are: 1) Cost reduction for internal processes like documentation and content generation / review 2) Time reduction for the execution of internal tasks performed by the employees, like generation of documents such as IFU (Instructions for Use of medical substances) I am also responsible for onboarding new members of the AI team and guide developers in their integration and immersion in Siemens Healthineers ecosystems. The main communication channels for those tasks were documentation of internal services and seminars for introducing technologies used in Siemens Healthineers. The main technologies used to develop the Applications are based on the Azure Cloud Provider, especially Azure App Services, Azure AI Search, Azure OpenAI, Azure Video Indexer and Azure Document Intelligence. The Application stack I developed on was composed by: - NodeJS, React, Vite, with JavaScript and TypeScript - API Development with Falcon, FastAPI, Flask and Quart (Python) - Event-Driven Architecture, using Asynchronous communication and choreography of microservices - Terraform, Azure Bicep and bash scripts for infrastructure deployment and server configuration and management To ensure that the solutions we develop are in alignment with the business domain, I used ITIL for ITSM. To ensure that data is safely stored and managed through the whole workflow of the solution, I used the COBIT framework. Keywords: Azure AI Search | Azure OpenAI | LLMOps | Observability | ITIL | ITSM | COBIT | TOGAF | Terraform | Azure Bicep | React | Node | Python | TypeScript
PhD research, where I focused on two different domains: planetary internal structure models and time series data analysis. For the first specialization area, I was focused on modeling internal structure of planets and stars, low Reynolds number hydrodynamics, and the effects of tidal interactions on the spin-orbit evolution of planetary systems. Developed numerical codes in C++, FORTRAN, Python, and Rust for orbital evolution calculations and solving the Navier Stokes Equation. For the second focus, that is, time series data analysis, I focused on Fourier Transforms, DFT and FFT, as well as periodogram analysis using Lomb-Scargle periodograms. I then moved to more advanced models for finding optimal solutions to temporal signals. The main techniques I developed were based on Monte Carlo Methods, more specifically BMC (Biased Monte Carlo) and MCMC (Markov Chain Monte Carlo). In all those cases, I used Python as the main language for doing chi-square minimization and finding the best-fit solutions for the time series. Regularization methods to prevent overfitting of solutions were also used to normalize models with different dimensions to a common score, thus allowing for an understanding of the best and more realistic solutions to model the data. Keywords: Shellscript | Python | FFT | Periodogram Analysis | C++ | Time Series Analysis | Regularization Techniques | Monte Carlo