About Nikolas de Hor
Nikolas de Hor is an AI Engineer based in São Paulo, SP, Brazil, working remotely. He focuses on LLMs in production: agents, RAG, and MCP (Model Context Protocol). Full stack is the complement, with Python, TypeScript, FastAPI, React, Supabase/PostgreSQL, Docker, and AWS.
How he works
Every engagement starts with a diagnosis of the problem, scope, constraints, and acceptance criteria. Deadlines and outcomes are agreed only after that. Corporate cases stay at a safe public level, without data, metrics, screenshots, internal names, or unauthorized operational details. Each public case separates problem, implementation, and verifiable proof (repository, documentation, package, or published product).
Capabilities
- Full stack: React, TypeScript, Node.js, Python, FastAPI, Flutter, APIs, integrations, and developer tooling.
- AI and agents: LLM applications, multi-agent orchestration, MCP, RAG, embeddings, semantic retrieval, OCR, transcription, computer vision, evaluation, and human fallback.
- Platform: PostgreSQL, MongoDB, Redis, queues, Docker, Kubernetes/OpenShift, CI/CD, observability, and AWS.
Experience
AI Engineer at CAST and lead architect of LUME, an enterprise hub for AI agents (RAG, per-project memory, model governance, cost and latency control), delivering for a client in the regulated banking sector. Maintains open-source projects under DeHor Labs, such as MCP Fiscal Brasil (45 tools, on PyPI and the official MCP registry), MCP Jurídico Brasil, and TranscreveAI.