Projects

A selection of experimental labs, production systems, and educational work. Links open the project detail page; the GitHub repo is inside.

Fante — RPG Game Orchestrator
FEATURED

Game-loop orchestrator for a voice-and-text RPG built for a 2-year-old. Coordinates LLM narration, game rules (via MCP), and player input through a ports-and-adapters architecture designed to grow with the player.

core-llm-bridge
FEATURED

Provider-agnostic LLM abstraction layer. One interface for Ollama, OpenAI, and Anthropic — sync, async, streaming — with conversation state, tool calling, and typed errors.

Copper — AI-maintained knowledge bases
FEATURED

An LLM Archivist reads raw sources — notes, PDFs, Obsidian vaults — and compiles a structured markdown wiki. No embeddings, no RAG; just folders, text, and LLM workflows (store / tap / polish). Supports multi-mind queries and bidirectional linking across knowledge bases.

FashionCLIP

Deep learning pipeline for fine-tuning multi-modal vision models on custom datasets. Implementation of Triplet Loss and custom metric definitions for semantic search.

PyTorch

Multi-modal Transformers

Fine-Tuning

Model Architecture

LoRA Fine-tuning

Hands-on implementation and analysis of Low-Rank Adaptation for efficient LLM steering.

nanoGPT from scratch

Educational reimplementation of autoregressive language models to master the fundamentals of modern Transformer architectures.

PyTorch

Transformer Training

Language Modeling

AWS_RAG

Serverless RAG system built for intelligent document processing and question-answering at scale.

Deep Learning Foundations Course

Comprehensive curriculum designed and taught at AFI Escuela de Finanzas. Includes advanced theory and practical implementations of Vision Transformers, Object Detection, and Generative Models.

HomeSync AI

AI-powered household management application leveraging multi-modal LLMs for document digitization and smart analytics.

Voice Cloning PoC

2-day research spike comparing voice-cloning routes — Azure Personal Voice, XTTS-v2, F5-TTS, and Qwen3-TTS — on a fixed evaluation grid. Key finding: fidelity-to-speaker and naturalness-in-target-language are in structural tension; automated metrics (WER/UTMOS) are blind to L1 accent transfer.

Translation Agents Lab

Experimental multi-agent framework for Machine Translation using open-source LLMs. Explores collaborative decision-making between models to surpass traditional translation benchmarks.

AI Agents

Open-source LLMs

RAG

Evaluation Frameworks