ANTONIO SERRANO / APPLIED AI ARCHITECT
I connect AI agents, data platforms and the economics of getting useful work done. Sometimes that means code. Sometimes hardware. Always a few more questions.
From compute and hardware to AI systems.
01 / THINGS TAKING SHAPE
Three connected questions.
Plenty of work still to do.
The request was cheap. Was the finished work? Exploring what AI actually costs once retries, data and human review enter the picture.
Research → prototypeCurrent foundation: FinOps specifications and vendor research. Live billing connectors and a measured workload baseline are still to come.
Next: One workflow. One cloud. One reconciled baseline.
Somewhere, the right dataset exists. An exploration of finding it across fragmented catalogues—with provenance, freshness and access conditions attached.
Active developmentThe local project includes catalogue discovery tooling and documentation, building on the upstream printing-press-library project. My specific contributions and bounded coverage are being documented for a public case study.
Next: A reproducible discovery benchmark and contribution record.
Upstream project ↗An agent needs a job description. And a stop button. Exploring one useful workflow with explicit permissions, human review and a way back when things fail.
Experiment designFabrec is an existing recruitment-data application and a possible test case. Autonomous completion and economic impact have not yet been demonstrated.
Next: Instrument one workflow, including exceptions and human review.
A project’s status is part of its story. Prototypes, plans and measured results are labelled separately.
THE RESEARCH SIDE OF THE WORKBENCH
Who’s building businesses that run themselves—and what still needs a person? My public research into the tools, operating boundaries and evidence behind the autonomy story.
Explore Autopilot Index ↗AI reliability, adoption and the gap between a real need and a business.
Read the field note ↗02 / THE PERSON BEHIND THE PROJECTS
My work has taken me from hardware and industrial electronics to cloud, data platforms and applied AI.
I’m interested in the joins: where an agent finds its data, what happens when it gets something wrong, and whether the outcome was worth the resources it used.
zetabyteLab is where I explore those questions through independent projects, research and practical experiments. A working notebook, with the unfinished bits left visible.
My professional background ↗03 / FOLLOWING THE CURIOSITY
Because not every interesting problem
starts with a prompt.
Hardware, mechatronics and new ways to interact with intelligent systems. Sometimes you need to touch the thing you’re building.
Helping a friend build a Basque cheesecake business. A very physical reminder that operations happen outside a browser, too.
Exploring a more personal language-learning experience through daily practice. One useful conversation at a time.
Enterprise agents, data platforms, AI economics—or a question that connects them. I’d like to hear what you’re working on.