Hey, I'm Anderson, a Senior Software Engineer from Brazil focused on AI agents and system architecture.
At Conty, I work on the platform that connects brands and creators, and on the AI agent that runs inside it. The interesting part is not the model, it's everything around it: turning a conversation into a real campaign with budget, briefing and payouts, keeping tool calls safe against a multitenant database, and measuring whether the agent actually saved anyone time. Most of my engineering effort goes into the boundaries, permissions, guardrails and failure paths, because that is where agents stop being demos and start being infrastructure.
That same interest shows up in what I build outside work. agentbridge lets you define an agent tool once and either call it in-process or serve it over MCP, so a single definition works in both worlds. llmtape records real LLM API calls once and replays them deterministically in tests, which makes it possible to assert exactly which tools an agent called instead of hoping the model behaves.
Recently I have been writing native Swift. speech.md is a macOS app for voice transcription that runs entirely on device: dictate into any app with a global hotkey, record meetings with your microphone and the system audio on separate channels, and transcribe audio files. No server, no account, nothing leaving your Mac.
My stack is mostly TypeScript with NestJS, Next.js, Cloudflare Workers and Postgres, plus Swift when the problem belongs on the machine. I care about software that holds up outside the happy path: clear boundaries, honest error handling, and the version that still makes sense six months later.
You can find more of my work on my portfolio, LinkedIn, X, and GitHub.