* PORTFOLIO · 2026 · REMOTE, BRAZIL

Victor
Dantas

I build software for millions. I research how AI agents should work with us.

CURRENTLY Software Engineer at Bamse · AI Committee & Mentorship
CONTACT victordantasdeveloper@gmail.com
01 About

I'm a mid-level full-stack engineer with more than four years building serious products in production — and I'm at the frontier of AI applied to engineering.

Today I work at Bamse, where I maintain microservices for Unilever's OMO Lavanderias ecosystem — over a million users across orders, bookings, payments, and notifications, with cross-platform clients on iOS, Android, Web, and Desktop.

In parallel, I'm part of Bamse's AI Committee, where we research the best methods for effective use of agents — context engineering, harness engineering, skills, MCP, hooks, and subagents — and I mentor in the AI Mentorship program, guiding client companies through the practical adoption of AI-augmented engineering workflows.

I have no interest in being "the dev who uses ChatGPT". I'm interested in building the tools, the contracts, and the habits that make agents genuinely useful — and in teaching the people adopting them.

02 Mentorship & AI Research

"Agents are only useful when the environment around them is designed with the same care we give to a public API."

* AI COMMITTEE · BAMSE

Research on agents

A living conversation on what works in production: context engineering, harness engineering, Claude Code skills, MCP, hooks, subagents. We translate findings into guidelines, templates, and reviews for the whole team.

  • — Context engineering applied to agents in production
  • — Harness engineering: the environment around the model
  • — Skills, MCP servers, hooks, and subagents as a contract
* AI MENTORSHIP · BAMSE

Mentorship for companies

I guide Bamse's client companies through the real adoption of AI agents — not demos, but engineering workflows that survive code reviews, deploys, and oncall.

  • — Diagnosis of the current engineering workflow
  • — Design of domain-specific skills, MCPs, and hooks
  • — Adoption follow-up and retrospectives

Thinking about adopting agents on your team?

Talk about mentorship
03 Experience
Jun 2025 — Presentongoing

Bamse

Mid-level Software Engineer · AI Committee · AI Mentor

Unilever's OMO Lavanderias ecosystem — 1M+ users. Microservices for orders, bookings, payments, and notifications; cross-platform clients on iOS, Android, Web, and Desktop. Code reviews and definition of engineering standards.

* AI COMMITTEE · AI MENTORSHIP

In committee, I research the best methods for effective use of agents — context engineering, harness engineering, skills, MCP, hooks, subagents — and I guide client companies in the practical adoption of these workflows.

Nov 2023 — Jun 20251 year, 7 months

Fox IoT

Software Engineer

Energy distribution platform: real-time operational dashboards, alerts, remote commands for field devices, and compliance with ANEEL (DRP/DRC).

Feb 2023 — Nov 20239 months

Softaliza

Full Stack Engineer

Scientific events platform with 100K+ attendees: on-site check-in, DOI flows, ~40% reduction in load time, and leading the migration from JavaScript to TypeScript.

Feb 2022 — Feb 20231 year

Softaliza

Front-End Engineer

Performance, automated testing, and initial TypeScript adoption on a scientific events platform.

04 Featured Projects
View all projects
BAMSE / UNILEVER · 2025 Nº 01

OMO Lavanderias

App ecosystem serving 1M+ Unilever users in Brazil — microservices for orders, bookings, payments, and notifications, with iOS, Android, Web, and Desktop clients.

NestJS ·React Native ·PostgreSQL ·RabbitMQ
Ver case study
SOFTALIZA · 2023 Nº 02

Scientific Events

Platform with 100K+ attendees: on-site check-in, DOI flows, ~40% reduction in load time, and a developer-led migration from JavaScript to TypeScript.

React ·TypeScript ·Node.js ·Cypress
Ver case study
PERSONAL · 2026 Nº 03

Resume Skills Toolkit

Claude Code skills that generate ATS-friendly résumés via Pencil + LaTeX + Tectonic — a declarative pipeline from .pen to typographically correct PDF.

Claude Code ·Pencil ·LaTeX ·Tectonic
Ver case study
PERSONAL · ONGOING Nº 04

MCP Server (ongoing)

Personal research into design patterns for genuinely useful MCP servers — beyond hello-world.

MCP ·TypeScript ·Claude Agent SDK ·Ongoing
Ver case study
05 Skills
* FEATURED CATEGORY

AI applied to engineering

*AI Agents*Context Engineering*Harness Engineering*Skills*MCP*Hooks*Subagents*Claude Code*Claude Agent SDK*Applied research on agents*AI mentorship

Languages & Runtimes

TypeScriptJavaScriptNode.jsSQL

Backend

NestJSFastifyREST / OpenAPIWebSocketMicroservices

Frontend & Mobile

ReactNext.jsReact NativeExpoElectronVite

Data & Infra

PostgreSQLRedis / RedlockRabbitMQDockerCI/CD

Architecture & Testing

DDDSAGAEvent-DrivenJest / VitestCypress
06 Recent Writing
May 14, 2026 AGENTS

Context engineering in practice: what I learned keeping agents in production

Most of the failures I've seen with agents in production didn't come from the model — they came from the context. Notes on what goes in, what stays out, and why.

8 min
Apr 2, 2026 MCP

MCP beyond hello-world: patterns for genuinely useful servers

Almost every MCP tutorial stops at an endpoint that returns the current time. The real work begins when the server has to handle auth, pagination, and errors transparently for the agent.

11 min
Mar 18, 2026 HARNESS

Harness engineering: why the agent's environment matters more than the model

The model is the engine. The harness is the whole car. Swapping models without touching the harness is like tuning the injection without checking the tires.

6 min
Read all writing
07 Contact

Recruiters, founders, CTOs, or teams curious about AI agents — get in touch.

EMAIL victordantasdeveloper@gmail.com LINKEDIN /in/victordantasdev ↗ GITHUB /victordantasdev ↗
Send email Book a chat