TechCafe

Breaking the Box: Beyond the Limits of Current AI

Modern AI is a powerhouse of productivity — but its capabilities are bounded by real engineering constraints. 'Breaking the Box' explores the strategies we use to refine model training, overcome current agent limitations, and leverage the unique strengths of LLMs. By understanding where pattern recognition ends and autonomous logic begins, this Tech Talk offers a grounded view of how to maximize AI's potential while maintaining rigorous standards of predictability and performance.

clock 6:00 PM - 8:00 PM | Thursday, May 21, 2026
address Dos Marias Restaurant
Address: C. Tapalpa 1, Vallarta Poniente, 44110 Guadalajara, Jal.                                      

 
 
 
 

Our Partners

katalon-logo-1536x391

Meet Our Speakers

Agenda

6:00 PM – 6:25 PM

Welcome & Opening

Topics & Speakers Introduction

6:25 PM – 6:45 PM

Topic: AI is powerful… but bounded

This talk explores the real limitations of modern AI models, explaining why LLMs don’t truly reason but instead predict patterns, and how their probabilistic nature impacts reliability. It also shows how to design systems that leverage their strengths while maintaining control and accuracy.

Speaker: Carlos Zuñiga | AI Engineer | KMS Technology

6:45 PM – 7:05 PM

Topic: From LLM Capabilities to Reliable AI Systems

Large Language Models are powerful, but raw capability alone doesn't make a reliable product. The real challenge lies in the gap between what an LLM can do in isolation and what it takes to make it predictable, dependable, and production-ready. This talk explores practical approaches to closing that gap: from evaluation frameworks and thoughtful system design to the controlled use of agents — so you can build AI systems that don't just impress in demos, but hold up in the real world.

Speaker: Arturo Rodriguez | AI Engineer | KMS Technology 

7:05 PM - 7:25 PM

Topic: Beyond the Hype: Designing Reliable AI Within Its Limits

This talk explores the real boundaries of modern AI systems, moving beyond hype to understand how large language models actually work. We examine their probabilistic nature, why they struggle with reliability and reasoning, and how to design systems that harness their strengths while mitigating their limitations.

Speaker: Kevin Zepeda | AI Engineer | KMS Technology 

7:25 PM - 7:55 PM

Panel Discussion

Speakers:

Carlos Zuñiga | AI Engineer | KMS Technology 

Arturo Rodriguez | AI Engineer | KMS Technology 

Kevin Zepeda | AI Engineer | KMS Technology 

7:55 PM - 8:00 PM

Closing

Thank you & Goodbye