~/about.md

Ricardo Silveira

Biologist (PhD), data scientist and developer.

I am a biologist with a PhD in Zoology and two postdocs, and I have worked in data and software development since 2020. For the last two years I led the Innovation area of an environmental consultancy as Product Owner and Tech Lead.

I code in TypeScript, Python and R. I work with clear scope, acceptance criteria and documented decisions, and I use generative AI in development with the gain measured.

Available from late November 2026. Looking for roles in data, product and applied AI, preferably remote.

Get in touch cat contact.txt

By the numbers

~45people using the time tracking system I led as PO

14 → 2internal systems consolidated

~70%less time on AI-assisted tasks, measured in logged hours

3–6kmonthly users of the chatbot I built

Selected work

internal product · Seteg Project and time tracking The project and time management system that replaced a paid tool and is used across the whole company. projects/gestao-de-horas.md app suite · Seteg Field data collection Apps that replace paper forms and spreadsheets in fauna and flora surveys, so records arrive structured instead of being retyped weeks later. projects/coleta-de-campo.md open science tetrapoDNA Which Brazilian amphibians, reptiles, birds and mammals still have no public DNA sequence at all? projects/tetrapodna.md applied AI banca-simulada Claude Code plugin that reviews a research proposal the way a thesis committee or grant panel would. projects/banca-simulada.md

How I work

Spec before code.: A project starts as a written contract: brief, domain research, then requirements split into small shippable phases, each with explicit success criteria.
Agents inside a process.: I drive Claude Code through custom skills for planning, execution, verification and review, so the same discipline applies whether the code comes from me or from an agent.
Memory with human review.: An Obsidian vault where a daily journal goes in and structured proposals come out. I review every one before it is written.
Passing tests are not proof.: Opening the published product and using it by hand is part of my definition of done.