Carle Medina

I’m a BS Computer Science student (expected May 2026) who builds systems that balance performance with real-world usefulness. My coursework in Data Structures and Algorithms and Database Management System keeps my engineering grounded in solid fundamentals.

I gravitate toward backend and data engineering because I enjoy turning messy inputs into reliable outputs. In a Tagalog Wikipedia Corpus Cleaning Pipeline, I built a 7-stage NLP preprocessing pipeline and reduced a 6.26M token corpus by 51% while preserving morphologically rich vocabulary, then validated quality using MSTTR.

I also love learning how modern AI actually works by implementing it end-to-end. In “Transformer / LLM from Scratch (PyTorch),” I implemented key transformer components—scaled dot-product attention, multi-head attention, causal masking, and residual connections—without high-level abstractions to understand modern LLM internals.

Beyond projects, I’ve led and shipped in collaborative settings through hackathons. I acted as Acting CEO for a DICT hackathon (React, Node.js, Express.js, MongoDB, JWT) and built the backend for BlueSphere during the NASA International Space Apps Challenge, integrating the EarthAccess API with Python to process satellite data into interactive learning tools.