Caio Rocha

Contributed to applied 3D generative AI within Dassault Systèmes' global virtual twin initiative, focusing on prototyping and delivering a production-ready MVP. Key Contributions: - Adapted a motion generation model from NVIDIA for a VR software in C++ - Trained 3D diffusion models for more realistic movement with a focus on customer requirements - Allowed integration into production by exporting a first-ever ONNX version of this model - Contributed to a FastAPI service integrating an LLM in a containerized setup - Restructured data for smoother downstream integration by a product team Core competencies: - PyTorch, FastAPI, Docker, Blender - 3D, Diffusion, AR/VR - PoC, MVP delivery - CI/CD, Sprints

CAEd/UFJF is a leading reference in large-scale educational assessment in Brazil. I worked in the AI Innovation team, focusing on building practical NLP systems for reading assessment. Key Contributions: - One of the main contributors of Brazil's first automated fluency assessment system. - Fine-tuned Transformer-based ASR models and implemented a pipeline handling terabytes of audio data. - Released a system that reduced data annotation time by ~80%, from weeks to days. - Built analyses and visualizations to support decisions with education experts. - Contributed to applied AI research and national conferences. Core competencies: - Stack: PyTorch, Hugging Face - ML: Speech recognition, NLP, data analysis - Software: Python, HTML, CSS, REST APIs - Agile: Scrum, Kanban - Visualization: Excel, Matplotlib

An international 8-week program for aspiring technology leaders from all disciplines, teaching them to use cutting-edge technology tools in data science and AI to tackle societal challenges. I benefited from a fully founded scholarship that provided me with cohort-based classes, mentorship, and networking opportunities.

Joint research, teaching, and mentoring scholarship under supervision of Prof. Elson Toledo, Prof. Diego Gomes, and Prof. Leonardo Goliatt. My research projects were entitled "Machine Learning for Classifying COVID-19 Drug Candidates" and "Water Flow Forecasting at the Cahora Bassa Dam through Deep Learning and Statistical Learning." I co-authored two papers published in international journals.