Felipe Euphrasio

PhD-level AI Engineer and Applied Researcher with expertise in Computer Vision, Generative AI, NLP, and Agent Orchestration, delivering AI solutions for real-world industrial and research challenges. Experienced in designing and deploying systems involving deep learning, LLMs, VLMs, multi-agent workflows, visual odometry, and data fusion, with applications in drone autonomy, oil and gas, agritech, healthcare, and intelligent automation. Combines hands-on technical depth in Python, ML/DL frameworks, vector databases, cloud environments, and geospatial data with a strong track record in technical mentoring, startup acceleration, and cross-functional leadership. Recognized for problem-solving, adaptability, clear communication, collaboration with multidisciplinary teams, and the ability to translate complex AI concepts into practical business and research outcomes.

• START DEEPTECH Bootcamp - Speaker, 2025, Certificate awarded by Sebrae-SP - Brazilian Support Service for Micro and Small Enterprises, in partnership with Wylinka and the Institute for Technological Research (IPT), recognizing participation as a speaker at the START DEEPTECH Bootcamp held on May 13-15, 2025, totaling 24 hours of professional training and knowledge exchange on innovation, entrepreneurship, and deep technology.
• 63rd Brazilian Congress on Medical Education (COBEM 2025), 2025, Certificate of presentation for the work 'Interprofessional Innovation: Using Melanoma Detection for Teaching Artificial Intelligence in Medical Sciences', authored by Pedro Carlos da Silva Euphrásio, Gabriel Barcellos Andrade Camargo, Felipe Pacheco de Almeida Euphrásio, Felipe Camargo Segreto, Mariana Ferreira da Silva, and Rinaldo Henrique Aguilar da Silva. The poster was presented at the 63rd Brazilian Congress on Medical Education (COBEM), held in Natal, Rio Grande do Norte, Brazil, from September 11 to 13, 2025.
• START DEEPTECH Bootcamp - Mentor, 2025, Certificate granted by Sebrae-SP, Wylinka, and IPT for serving as a mentor during the START DEEPTECH Bootcamp, held from May 13 to 15, 2025, with 24 hours of mentoring and technical support to innovation-driven startups in the field of deep technologies, guiding multidisciplinary teams in project validation and strategic development.
• Start DeepCamp 2024 - Mentor, 2024, Certificate granted by Sebrae-SP, Wylinka, and the Institute for Technological Research (IPT) for serving as a mentor at the START DEEPCAMP 2024, held from September 24 to 26, 2024, in São Paulo. The event focused on guiding innovation-driven startups through mentoring sessions on artificial intelligence, data science, and technology-based entrepreneurship.
• Spark Camp 2024 - Innovation & Technology Program, 2024, Certificate of participation in Spark Camp 2024, an immersive program focused on innovation, entrepreneurship, and deep technology development. The event gathered researchers and industry professionals to collaborate on practical challenges in AI, data analysis, and automation.
• IV Scientific Meeting - Faculty of Medical Sciences of São José dos Campos (Humanitas), 2024, Certificate awarded for receiving Honorable Mention for the research work 'Biomedical Engineering: Concepts and Practices of Artificial Intelligence Techniques for Melanoma Detection', authored by Gabriel Barcellos Andrade Camargo, Felipe Camargo Segreto, Mariana Ferreira da Silva, Pedro Carlos da Silva Euphrásio, and Felipe Pacheco de Almeida Euphrásio. The work was recognized at the IV Scientific Meeting of the Faculty of Medical Sciences of São José dos Campos - Humanitas, held from November 28 to 30, 2024.
• XXII Workshop on Applied Computing (INPE - National Institute for Space Research), 2022, Certificate of presentation at the 22nd Workshop on Applied Computing (WorCAP 2022), organized by the Graduate Program in Applied Computing (PGCAP) at INPE - National Institute for Space Research, on September 15, 2022. The presented work, titled 'Use of Convolutional Neural Networks in Vehicle Detection and Re-identification from Multiple Drone Images', focused on the application of deep learning and computer vision techniques for aerial image analysis and multi-drone coordination.

My experience spans a wide range of AI applications across industry, research, and innovation. In Computer Vision, I have worked on projects involving fraud detection in dental images, driver drowsiness estimation through blink and yawn monitoring, large-scale bovine facial recognition, multispectral fusion of satellite imagery for agriculture and aquaculture, failure detection in robotic arm assemblies, drone position estimation through visual odometry using visible and thermal imagery, detection and re-identification of suspicious vehicles, pothole detection on highways using drones, melanoma diagnosis from medical images, horizontal road marking detection in Street View images, agricultural pest detection from mobile phone images, sugarcane health monitoring through computer vision, oil spill and marine debris detection, warehouse inventory counting with drones, and farm meter digitization using OCR and image preprocessing.

To support these solutions, I have applied deep learning and vision models such as YOLO, U-NET, GANs, ZoeDepth, MiDaS, SuperGlue, LoFTR, EasyOCR, Tesseract, FaceMesh (MediaPipe), and LSTM.

In Machine Learning and Data Science, I have developed solutions for regression modeling across different domains, prediction of customs and maritime shipping stages in import/export processes, forecasting major cardiovascular events from medical history, and breast cancer prediction using clinical and imaging data. My toolkit includes algorithms such as SVM, Random Forest, Decision Tree, XGBoost, PCA, and K-means.

In Generative AI, NLP, and Agentic Systems, I have contributed to multi-agent systems for interpreting legislative documents, LLM-based pipelines for data-driven insight generation, LLM integration for real estate recommendation based on user profiles, interactive chat systems powered by LLMs, and multi-agent orchestration for specialized workflows across multi-domain problems. In this context, I have used technologies such as Groq API, Ollama, RAG pipelines, VLMs including Kosmos, LLaVA, BLIP, and Qwen, as well as LLMs such as OpenAI models, Llama, DeepSeek, and Mistral.

My technical stack also includes platforms and frameworks such as AWS SageMaker, Ollama, Hugging Face, n8n, QGIS, Qdrant, Pinecone, ChromaDB, Amazon RDS, PyTorch, TensorFlow, Transformers, Pandas, Matplotlib, NumPy, LabVIEW, C, and Python. I also work with geospatial and remote sensing data, including Sentinel-2, Landsat, PlanetScope, and Street View imagery.

From a delivery and collaboration standpoint, I am familiar with Kanban, Agile methodologies, and Lean Startup principles, combining technical depth with a practical approach to problem-solving, experimentation, and innovation.