Final year CS student at UFV. Currently tackling several personal projects, including the design, implementation and training of a novel non-autoregressive language model architecture.
- Co-author of "Benefits of Chain of Thought Prompting for Clinical Record Rubric Evaluation in Undergraduate Medicine Education", accepted at JMIR Medical Education (2026, JIF 12.6, #1 of 86 in "Education, Scientific Disciplines"). CoT prompting on Llama 3.1 and Claude 3.5 raised evaluation accuracy from ~86% to 94.6% and surfaced 317 errors in the human-graded rubrics: doi.org/10.2196/88652
- Advanced knowledge of large language models (LLMs): deep understanding of architecture, training dynamics and fine-tuning methods.
- Deep Learning w/ PyTorch: designing, training & deploying neural networks for vision, language and multimodal tasks.
- Neural Network Training Infra: currently using several cloud computing platforms for my CS final project.
- Hands-on experience with agentic platforms: using Claude Code, Codex, Jules (among others), both sync (active supervision) and async (triggered execution) for several projects.
- Long-time GenAI user: image models since Stable Diffusion 1.4, having implemented ComfyUI node workflows for custom pipelines.
- DevOps: Reproducible environments & deployments using Docker / Compose.
- Web/App development: Several shipped personal projects using Next.js and React Native (Expo).
- Image Generation Models: diffusion models, RFs, and related techniques.
- Reasoning Models: architectures and methods for complex, multi-step problem solving.
- Neural Architecture Design: custom architectures tailored to task-specific goals and constraints.
Research collaborations and roles where I can keep working at the intersection of generative modeling and applied research.
I want my work to have a real-world positive impact, I'd love to contribute to medical applications, education, or other socially beneficial domains where AI can make a difference.
Most of my previous and ongoing work is private; happy to walk through it in a conversation.
Raise an issue, start a discussion, or find me on: