Juan Manuel Ortiz de Zarate

Profile

Join date: May 16, 2024

About

0 likes received

0 comments received

0 best answers

I hold a Bachelor's degree in Computer Science from the University of Buenos Aires (UBA) and am about to defend my doctoral thesis focused on AI. I have taught Artificial Intelligence at the Faculty of Engineering of UBA and am the Co-Founder and CTO of Data Voices, a company dedicated to the development of AI technologies. Additionally, I worked alongside Peguy Luyindula, a former player of the French national team, in developing an innovative AI application for football contract negotiations.

Badges

AI Expert
Worked as a Data Scientist/AI Expert and/or is an AI Researcher

Overview

Location

Buenos Aires, Argentina

Posts (67)

Dec 18, 2025 ∙ 10 min

What If Reasoning Doesn’t Need Billion-Parameter Models?

Large language models excel at language but often struggle with structured reasoning tasks. This article explores Tiny Recursive Models (TRMs), a radically simpler approach that uses small neural networks with recursive refinement to outperform massive LLMs on puzzles like Sudoku, mazes, and ARC-AGI. By prioritizing iterative reasoning over scale, TRMs show that deep thinking can emerge from minimal architectures, challenging prevailing assumptions about model size and intelligence.

Dec 13, 2025 ∙ 11 min

Teaching Robots to Dance

RoboBallet explores a new approach to multi-robot task and motion planning by combining graph neural networks with reinforcement learning. Instead of decomposing planning into brittle subproblems, the system learns to coordinate multiple robotic arms directly through structured relational reasoning. Trained in simulation and generalizing zero-shot to real workcells, RoboBallet demonstrates how learning-based coordination can scale to industrial environments where classical planners fail.

Dec 6, 2025 ∙ 9 min

When Models Learn to Think Before Painting

This article explores HunyuanImage 3.0, Tencent’s groundbreaking open-source multimodal model that unifies language understanding, visual reasoning, and image generation. It examines the model’s data pipeline, architecture, Chain-of-Thought workflow, and progressive training strategy, showing how HunyuanImage 3.0 achieves state-of-the-art text-to-image performance while enabling richer control, coherence, and creativity.