AI-Native Backend Engineer

I build AI systems that run in production.

For the past two and a half years I've been at Pedbot, where I helped build a conversational AI product from its first version to enterprise deployments at Boehringer Ingelheim and DPSP, one of Brazil's largest pharmacy chains. It handles around 5,000 customer conversations a day. I designed the multi-agent architecture with LangChain and LangGraph, built the RAG pipelines that ground answers in client documents, integrated Whisper for voice input, and wrote the WhatsApp Cloud API and CRM integrations. The backend is NestJS, TypeScript and Python over PostgreSQL and Redis.

The result I'd point to: I concluded our agent architecture wasn't going to scale, argued for a rewrite instead of a patch, and delivered it. Inference cost came down 85% and response latency by more than half, in production.

What sets me apart is the operational side. I started out in technical support, reading logs and writing SQL in other people's systems, so I diagnose LLM failures by tracing them: whether a wrong answer came from the prompt, the knowledge base, retrieval, or the model itself. I've also built internal agents with Claude Code that pull production bugs out of our logs and open the issues to fix them.

I'm looking for remote work where AI is the core of the product rather than one feature of it, with ownership end to end. Fluent English, based in Brazil (UTC-3), fully remote since 2024.