Thiago BertoldiEngineering / Personal essay
Overcoming the Urge to Over-Engineer: My Blog Stack
On wanting a place to write, wanting it to look cool, and the elaborate detour between those two things.
Engineering human–AI collaboration
Design.
Build.
Question.
I’m Thiago. I design workflows for people and AI agents, and build the infrastructure that puts AI to work.
Clear responsibilities. Context that travels with the task. Evidence to decide what happens next. That’s the kind of engineering I care about.
Explore my work
Build with intention.
Stay curious about the result.
Share what you learn.
BuildingExpert Software Engineer · SUSE AI
InvestigatingMSc student · UFSC
SharingSUSECON · SAAC / UFSC
A workflow worth designing
- 01Define the task
- 02Delegate with context
- 03Inspect the work
- 04Decide what ships
01 / Writing
Notes from the work.
AI infrastructure / Technical article
Observability for AI Workloads: A Primer
An introduction to the signals, infrastructure, and operational questions around generative AI.
02 / Selected work
Things I help build.
Things I build for myself.
SUSE / Contributor
SUSE AI Factory
Contributions spanning workload deployment, Fleet integration, and private application sources.
Explore the implementation: Fleet delivery · Private sources
Kubernetes / Fleet
SUSE / Development lead
SUSE AI Observability Extension
Turning GenAI telemetry into topology, dashboards, and health monitors. I also contribute to the AI OpenTelemetry Collector.
Kubernetes
Personal / Local AI
smarty-pants
Select text. Press a shortcut. Rewrite it with a local model, from your Wayland desktop.
Writing tools
Personal / Desktop and workflow
A desktop that feels like mine
Hyprland on openSUSE, bringing writing, study, music, and terminal tools together.
Hyprland
03 / Research in progress
What belongs
around a model?
My MSc research at UFSC explores how software-task specifications and organizational experience can guide the selection and adaptation of coding-agent models and harnesses.
Tools, instructions, context, and controls all belong in that question. So does the person deciding whether the result is useful.
MSc in Knowledge Engineering and Management · Ongoing
The question I’m working on
01
Understand the taskRequirements, context, constraints
02
Configure the agentModel + tools + instructions + controls
03
Review the outcomeEvidence, human judgment, reusable experience
04 / Speaking
Let’s make it understandable.
SUSECON2026Tutorial / English
AI Observability with OpenTelemetry: Patterns and Best Practices
Co-presented with Gunasekhar Matamalam.
Watch the sessionSAAC / UFSC2026Talk / Portuguese
Agentes de IA na Engenharia de Software: Pensamento Sistêmico além das amarras da linguagem
Agent anatomy, context, and task decomposition: designing the work around the model.
Explore the slides Session recap ↗SAAC / UFSC2025Talk / Portuguese
IA Generativa Open Source
Privacidade, segurança, liberdade e comunidade. Da demanda organizacional ao desenho de uma aplicação com IA.
Download the slidesOriginal 2025 deck · PDF, 29 slides · 1 MB
Planning an event on agent workflows, AI infrastructure, or observability? Let’s talk.
05 / A little more context
The person
behind the terminal.
I came to AI infrastructure through control and automation engineering, embedded Linux, IoT, and data platforms. I’ve led engineering teams. Today, I also design and direct AI-assisted workflows. Clear expectations, useful context, and thoughtful review keep turning out to matter.
I think knowledge becomes more useful when it is shared. That connects my teaching with open source, digital sovereignty, and technology that helps people exercise their judgment.
I also enjoy making my desktop feel right and placing an elaborate word somewhere you weren’t expecting it.
Experience and education