Hi, I'm Victor.
Software engineer, full-stack, more than four years building production products. Today at Bamse, I live between two worlds that reinforce each other: maintaining Unilever's OMO Lavanderias ecosystem for a million users — and researching, in committee, how AI agents should work with us.
I started on the front end, working on CSS for scientific event plans at Softaliza in 2022. The following year I migrated the entire codebase to TypeScript. I crossed over to full-stack in 2023, and in 2024 I joined Fox IoT, where I learned to respect systems that have to work in real time for energy infrastructure.
In 2025 I joined Bamse to work on the biggest project of my life — and almost in the same month, the team formed the AI Committee. It was the best timing I could have asked for: learning product and agents at the same time, at real scale.
Today I split my time between production code, applied research in the Committee, and mentorship for client companies adopting agents.
Domain first, code second.
Good code is a consequence of understanding the problem. I always try to spend more time with product, operations, and real users before proposing a solution.
I work in vertical, observable slices.
Each PR solves a problem end to end — and leaves a visible mark in logs, metrics, or behavior. Big-bang is the most expensive way to learn you were wrong.
Agents are colleagues, not shortcuts.
I don't delegate understanding to the model. I use agents to speed up specific mechanical parts of my flow — and I hold myself accountable for the results as if they had been written by hand.
Mentorship is part of the work.
Knowledge that doesn't circulate becomes a bottleneck. Technical and generous code review, clear ADRs, and explicit mentorship are part of the product — not an extra.
I'm part of Bamse's AI Committee and a mentor in Bamse's AI Mentorship program.
What I research
Context engineering, harness engineering, skills design, MCP servers, hooks as contract, and subagent orchestration patterns. Material applied in production, then systematized into guidelines for the whole team.
How I mentor
I work with client teams in short cycles: diagnosis of the current flow, design of domain-specific skills/MCPs/hooks, and adoption follow-up. Mentorship is less about demos and more about habits.
— Maintaining critical services for Bamse's OMO Lavanderias ecosystem.
— In active research at the AI Committee: patterns for MCP servers in production and guardrail hooks.
— Mentoring two client teams in practical adoption of skills and subagents for engineering flows.
— Building the personal Resume Skills Toolkit — Pencil → LaTeX → PDF pipeline.
Let's talk.
About engineering, AI agents, mentorship. About any of the three — or all three.