Mateo Torres
Arcade.dev
Rio de Janeiro, Brazil
Actions
Mateo is an expert at bridging cutting-edge AI research with practical developer tools. With a background in computational biology and experience in both research and engineering, he now focuses on making LLMs more useful by connecting them to systems developers already use. At Arcade, Mateo creates tutorials, SDKs, and open-source examples that help developers move beyond chatbots and into fully agentic applications.
Links
Area of Expertise
Topics
Tool Ergonomics: Designing MCP Tools an LLM Won't Misuse
Give an LLM ten tools and it will find creative ways to call the wrong one, pass a malformed argument, or loop forever retrying a call that was never going to work.
The model isn't reading your docs, it's reading your tool names, your parameter descriptions, and your error messages, and treating all three as prompts.
Shipping MCP servers across a range of integrations, I've watched the same design mistakes tank agent reliability long before the model's reasoning is the bottleneck. This talk is about the layer most people ignore: the shape of your tools.
I'll cover concrete patterns, when to split one tool into three versus collapse three into one; writing parameter descriptions the model actually obeys; returning errors that teach the model how to recover instead of retrying blindly; designing enums and defaults so an invalid call is impossible to construct; and how to test tool ergonomics without burning a fortune on full eval runs. Everything is framework- and SDK-agnostic and applies to any MCP server in any language.
Don't Delegate the Learning to the AI (Unless You Want to Get Dumber)
If an AI can write the essay, fix the bug, and explain the concept back to you, it's fair to ask why you'd bother learning any of it yourself. A lot of students and early-career engineers are asking me exactly that and losing motivation, because the superficial answer is "Learning is pointless now"
I want to give you a better answer. There's a difference between handing the AI a task and handing it your understanding. Let it do the task and you move amazingly fast. Let it do the understanding and you become someone who can't tell when it's wrong, can't fix it when it breaks, and can't grow past whatever the tool can already do. Aviation learned this the hard way decades ago: the better the autopilot got, the more dangerous it became to have a pilot who'd stopped really flying.
I'll show the difference between AI that accelerates your learning and AI that replaces it, why embracing the struggle of learning is still the best way to invest in yourself. And how to approach AI as the amazing tool that it is.
The Ambient Agent: Graduating Repetitive LLM Tasks into Deterministic Workflows
Most "agentic" systems fail in the same place: the agent is both the planner and the worker. Every run, it re-decides which API to call, in what order, with what arguments. That's expensive, slow, non-deterministic, and means nothing happens unless the user is steering the agent.
I built a multi-user YouTube analytics platform on the opposite premise. The agent is the configuration interface — it sets up channels, schedules, and notifications through MCP. The work itself runs as deterministic workflows triggered by cron. Configure once, run autonomously, query anytime.
This talk covers the architecture and the patterns that make it production-safe: a concrete heuristic for when a task should graduate from "agent decides" to "workflow executes"; a dual-MCP split that separates the agent's "hands" from its "brain interface"; and workflow patterns: parallel steps, typed step chaining, background execution, auto-cascade.
You'll leave knowing which parts of your agent are improvising when they should be on rails, and how to demote them without losing flexibility.
Architecting Reliable Agents with MCP Gateways
An agent is only as powerful as the tools it can reach, but as we move from simple prototypes to production systems, "MCP Sprawl" becomes a critical bottleneck. Directly connecting agents to a dozen disparate servers is fragile, insecure, and unmanageable at scale.
In this session, we explore the MCP Gateway pattern. A centralized layer that bridges the gap between non-deterministic LLMs and mission-critical systems. We will dive into the first principles of gateway architecture, from curated tool composition to "On-Behalf-Of" authentication. You’ll learn how to avoid common pitfalls like namespace collisions and over-privileged service accounts. Most importantly, we’ll demonstrate how to evaluate tool selections using relevant scenarios to ensure your agents remain reliable, secure, and vendor-agnostic.
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top