Machine Learning Engineer &
Windows application developer

Specializing in Speech Synthesis, Transformers, and C++ software dev.

View My Work

About Me

I started my journey with Visual Basic and C# as a child, moving to C++ in 2013. My early projects involved videogame modding (2017-2019) using C++, Qt, and WinAPI.

In 2020, I pivoted to Machine Learning, driven by a passion for speech synthesis. I went from zero Python experience to becoming a maintainer of TensorFlowTTS in 2021, where I contributed the first open-source TTS model capable of 44.1KHz audio output.

I briefly attended university (2019-2021) but dropped out to focus entirely on ML research and development. I built TensorVox, a C++ ML TTS application, and freelanced training/modifying models. In 2023 I joined a startup and started getting deep into model stuff

Today, I focus on Transformers for GenAI (Speech, Image, Text). I have experience finetuning GPT-J-6B on TPUs back in 2021 and hold a ROCm Star certificate as a result of my work with AMD hardware.

"I am not designed to come second or third. I am designed to win."

— Ayrton Senna

Buenos Aires, Argentina

Technical Arsenal

Core & Systems

C++ 90%
Qt 70%
WinAPI 65%
Linux 40%
C# 25%

Machine Learning & AI

Transformers 80%
ML (General) 75%
GANs 70%
TensorFlow 17%
Diffusion 25%

Hardware & Acceleration

HIP/ROCm 65%
TPUs 25%
CUDA 40%

Selected Projects

Echolancer Waveform
2025

Echolancer

A text-to-speech model trained on concatenated text-audio tokens using a modern LLM-style decoder-only approach. Features a 50 tok/s codebook for efficient audio language modeling.

View Demo GitHub
MQGAN
2025

MQGAN / MusicLSTM

Experimental audio tokenization using a spectrogram codec. Implemented a ResNet compressor and U-Net refiner for high-quality reconstruction.

Project Site Read Article
ReLUGT
2024

ReLUGT Activation

Proposed a parametric ReLU activation function inspired by KANs. Demonstrates faster convergence than SwiGLU in early tests.

Theory GPT Test
WinDiffusion
2024

WinDiffusion

A lightweight C++/Qt frontend for Stable Diffusion. No Python dependencies required for the end-user. Supports txt2img, img2img, and inpainting.

GitHub
Tacotron 2
2023

Tacotron 2 Conv Attention

A novel approach to solving attention stability issues in Tacotron 2 using Convolutional Attention Consistency.

Read thing
TensorVox
2021

TensorVox

High-performance local TTS application written in C++/Qt. Supports PyTorch and TensorFlow models. Widely used for freelancing deliverables.

GitHub
ZMapCharter
2019

ZMapCharter

Statistical map charting tool built with C++ and SFML. Supports custom shapefiles and CSV/XLSX imports.

Download
ZDEditorRS
2018

ZDEditorRS

Modern C++ rewrite of the database modding tool for SuperPower 2. Became the community standard for modding.

GitHub

Open Source Contributions

2021

44.1KHz Pretrained Model for TensorFlowTTS

Trained and released the first 44.1KHz open-source TTS model for the repo, introducing Multi-Band MelGAN + HiFi-GAN discriminator.

View PR
2020

C++ Support for TensorFlowTTS

Implemented C++ support for exporting and using TTS models via TensorFlow SavedModel C API.

View PR