Vinicius Arruda Email & Phone Number

Halifax, Nova Scotia, Canada

• Conducted Research and Development of NLP and LLM based solutions for understanding, managing, and generating requirements engineering.• Developed proof-of-concept products utilizing Large Language Models and Natural Language Processing techniques such as POS Tagging and Dependency Parsing.• Conducted analysis and parsing of source code utilizing parsing tools, NLP techniques and LLMs.• Developed prompts for private and open-source LLMs (GPT-3.5, GPT-4, Falcon 7B, Falcon 40B, Mistral 7B, LLaMA 2 7B, LLaMA 2 13B).• Developed solutions utilizing Python, OpenAI API, Hugging Face Transformers, Ollama, SpaCy, NLTK.

Campinas, São Paulo, Brazil

- Member of the Text-to-Speech (TTS) team.- Working on the entire text-to-speech (TTS) pipeline (text normalization, grapheme-to-phoneme, part-of-speech tagging, acoustic model, and vocoder) for Samsung’s embedded and cloud solutions for the pt-BR language.- Studying, training, and evaluating models such as Hidden Markov Model (HMM), Decision Tree (CART), Tacotron, and LPCNet.- Reviewing and developing normalization rules (Finite State Automata) for text normalization for the pt-BR language.- Actively interacting with teams from other countries such as Poland, South Korea and Greece.- Building corpus and lexicon.- Working with text anonymization with Named Entity Recognition (NER).- Main technologies used: Python, PyTorch, TensorFlow, C++, Docker, Perforce, Git, Jira, SpaCy, NLTK.

Vitória, Espírito Santo, Brasil

- Build and deployment of deep learning models in production.- Working with computer vision models, managing multiple cameras concurrently via inter-process communication.- Training and evaluation of object detection models for real-time inference.- Testing tiny models with Jetson Nano.- Integration of inference engines with back-end and front-end.- Front-end development with Angular.- Main technologies used: Python, PyTorch, TensorFlow, OpenCV, Pillow, gRPC, requests, JavaScript, TypeScript, Node.js, Angular, Docker.

Vitória, Espírito Santo, Brazil

- Extensively utilized PyTorch and TensorFlow frameworks for developing and optimizing deep learning models.- Explored and implemented advanced techniques in object detection, neural style transfer, synthetic data generation, and unsupervised domain adaptation.- Conducted cutting-edge research at the intersection of computer vision and domain adaptation, with practical applications in improving model performance across diverse visual datasets.

Vitória

Implementation and testing of image processing algorithms and convolutional neural networks applied to stereo images of cattle, in order to measure the cattle weight without the use of weight scale. During the internship it was used the programming languages Python and C++ with the libraries OpenCV and TensorFlow.

Vitória, Espírito Santo, Brazil

Research on artificial neural networks and metaheuristics, applied to time series prediction, focusing on the algorithms Extreme Learning Machine and Particle Swarm Optimization. This research resulted in a publication of an article at the XLVIII Simpósio Brasileiro de Pesquisa Operacional and was funded by brazilian research agency FAPES. The programming languages used during the research was MATLAB, C, C++, python and MQL5.

Master Of Science - Ms, Computer Science, 9.38/10.0

Computer Vision and Deep Learning

Bachelor'S Degree, Computer Science, 9.06/10.0

Técnico Em Informática

Bachelor'S Degree, Civil Engineering

Interrupted