I´m 23 years old, a Colombian technology enthusiast, self-taught fullstack developer dedicated to create tools that help people. while most people see daily obstacles as problems, i see them as challenges waiting to be solved.
I´m an active researcher in my university leading the development of 2 important AI projects which you could know about below in the notable projects section.
Currently I´m in the final stage of my university studies and i´m ready to build a strong professional relationship by contributing to high impact projects.

Developed a gateway to centralize all requests and secure backend microservices. It includes a custom Authentication and Authorization service that generates JWTs with an SHA-256 cryptographic key.

• Token Management: Tokens have variable expiration times (from 5 hours to 15 days), and users must reauthenticate after expiration to maintain security while minimizing inconvenience.
• Traceability: The gateway logs all activity in a database for monitoring and traceability.
• Security: The cryptographic key is stored in environment files specific to each server, ensuring unique local configurations and preventing unauthorized access.
• Deployment: The solution is fully containerized and deployed using Docker.

Implemented a convolutional neural network (CNN) to recognize Colombian Sign Language (LSC) from video feeds captured by six cameras, each positioned at a different angle.

• Dataset Creation: The system processes multi-angle video feeds to generate a compact dataset for training a sign-language translation AI.
• General Applicability: Although designed for LSC, the approach can be adapted for other regional sign languages.

Developed a Unity terrain generator system powered by a Conditional GAN (CGAN).

• AI-Generated Input: The CGAN produces 512x512 grayscale images, where pixel intensity corresponds to height percentages for terrain generation.
• Dataset: The AI was trained using Digital Elevation Maps (DEMs) from Colombian geography, provided by the United States Geological Survey (USGS).
• Technical Challenges:
  • Computational cost during training, mitigated by experimenting with Google Colab TPU (free), NVIDIA GPUs on Windows, and Linux environments.
  • Preprocessing DEM images in .tiff format, including converting grayscale gradients to global elevation points.
• Integration: Unity processes the grayscale images to generate realistic terrains with varying elevation levels.