With over 5 years of experience, I specialize in building scalable backend systems and deploying end-to-end AI solutions. My expertise lies in solving complex problems, implementing cutting-edge deep learning models (including LLMs), and delivering secure, production-ready systems.

• image-to-text-transformer: An end-to-end image captioning model utilizing Vision Transformer (ViT) and transformer architectures for generating descriptive text from images.

• text-translation-transformer: This is kind of an implementation of the original "Attention is all you need" paper, but translating english to portuguese.

• anime-recommender: A simple synopsis-based recommender built with FastAPI and a transformer encoder. Embeddings are projected with BigBird-ROBERTA (encoder), supporting k-Nearest Neighbors (kNN, L2 distance) and Support Vector Machine (SVM) for recommendations over 24 thousand animes.

• pytorch-ocr: A PyTorch-based Optical Character Recognition (OCR) small framework designed for accurate text extraction from images using CNNs, LSTMs and attention mechanism (no transformer, just the original attention modules).

• image-captioning: A deep learning project focused on generating textual descriptions for images using older approaches (cnn + lstm).

• pix2pix: Implementation of the Pix2Pix model for image-to-image translation tasks, enabling applications like style transfer and image synthesis. Quite outdated, but GANs were really something back then.

• yolov1: A PyTorch implementation of the YOLOv1 object detection model paper, capable of real-time object detection in images and videos.

• pytorch-object-detection-fundamentals: A PyTorch implementation of a classifier combined with a single object detector model, demonstrating fundamental concepts in object detection for beginners.

• server-client: A basic server to N clients in C sockets I've made for fun while I was in a vacation to show students how simple their life was with Python.

• aiyume: A platform for anime roleplay with LLMs that have long-term memory. I've made this project to improve my frontend skills (react and typescript) and to learn Stripe payment integration and Supabase (to have a reliable cloud sql db that wouldnt just break because its only living inside a container). Everything there was done by myself so its a pretty good baseline to have an overview of the stacks I work with.

  aiyume_preview.mp4

I've worked on a few other projects too but they may be too much outdated nowadays.