AI Engineer · LLM Agents · RAG/GraphRAG · Backend & Cloud

Founder @ Dewtech · Belo Horizonte/MG · Open to remote

I take AI prototypes from notebook to production. 10+ years in software engineering, now focused on LLM applications: multi-provider agents (OpenAI, Anthropic Claude, Gemini), RAG and GraphRAG architectures (Neo4j + Qdrant), prompt engineering and tool calling — with a strong backend foundation (NestJS, Python, .NET, Java, Rust) and Kubernetes (OKE) so things ship with observability and controlled cost.

Currently building the Fermio Platform — a multi-tenant AI agents platform — and the Fermio Engineering Copilot, applying GenAI to mechanical engineering (parametric CAD + FEA/CFD).

• LLM Applications — agents, RAG, GraphRAG, tool calling, structured outputs
• Vector / Graph Search — Qdrant, Neo4j, ElasticSearch vector, semantic search
• Agentic Engineering — author of the DARE Method (Design, Architect, Review, Execute)
• Production AI — multi-tenant, multi-provider routing, observability, cost control
• Cloud Native — Kubernetes (OKE), GitOps, multi-cloud (AWS, OCI)

Open-source Model Context Protocol server bringing YouTube intelligence into Claude Desktop — 18 production tools across search, video, channel analytics, benchmark, competitor research and content heuristics (CTR potential, retention signals, hook angles, content calendar). Backed by the official YouTube Data API v3 + YouTube Analytics API with OAuth2 (Brand Account support). Secure by design: AES-256-GCM token encryption at rest, SSRF guard, rate limiting, audit logging, Zod-validated inputs — all mapped to OWASP Top 10. Listed on the official Anthropic MCP Registry as io.github.dewtech-technologies/tubemind-secure-mcp. Install with npx tubemind-secure-mcp. → tubemind.dewtech.tech · github.com/dewtech-technologies/tubemind-secure-mcp

Open-source Model Context Protocol server connecting Claude Desktop to Obsidian — 6 tools (read/list/create/edit/delete/search notes) with Zod validation and winston audit logging. Built from scratch with OWASP Top 10 controls: path traversal protection, input sanitization, API key via .env only, localhost-only enforcement, size limits. Listed on the official Anthropic MCP Registry as io.github.dewtech-technologies/obsidian-mcp-secure. Install with npx obsidian-mcp-secure. → github.com/dewtech-technologies/obsidian-mcp-secure

Multi-tenant AI agents platform (monorepo pnpm, NestJS 11 + Nuxt 3). Multi-provider routing across OpenAI, Anthropic Claude and Gemini. Apps: api-agent-ai, orchestrator-api, chatbot-builder, atendimento-omnichannel, video-translator, synapse-process-audio (FastAPI) and code-assistant-vscode.

Generative AI for mechanical engineering. GraphRAG (Neo4j + Qdrant) over technical standards, parametric CAD generation with CadQuery/OpenCascade (STEP/STL), FEA (CalculiX) and CFD (OpenFOAM) pipelines. FastAPI + Nuxt 3 + Three.js (JWT-authenticated STL viewer).

Open-source toolkit for Agentic Engineering implementing the DARE Method (Design, Architect, Review, Execute). Single-package CLI with official-scaffold bootstrap (Composer, NestJS, Vite, Cargo, FastAPI, Go) — optionally via Docker images when the host has no toolchain — parallel DAG Task Runner (Kahn's algorithm + live canvas), mandatory Ralph Loop (build/test/lint runs per task before DONE), embedded GraphRAG engine (SQLite/JSON/Neo4j HTTP) and an MCP Server for deep IDE integration (Cursor, Claude Code, Antigravity). 8 stacks supported: Rust/Axum, Node/NestJS, Python/FastAPI, PHP/Laravel, Go/Gin, Go/stdlib, React, Vue + MCP servers (TS/Python). Install with npm i -g @dewtech/dare-cli. → github.com/dewtech-technologies/dare-method

Dedicated GitOps repo with Kubernetes manifests for Oracle Kubernetes Engine (OKE) — 12+ production apps, namespaces, secrets, Ingress Nginx, and infra (Postgres/Redis/Qdrant/Neo4j/RabbitMQ).

Rewrote a Python scoring API in Rust (Axum/Tokio) with six heuristic modules — no heavy deps (no OpenCV/NumPy). Major throughput gain and memory reduction in production.

If you're building LLM-powered products, agentic systems or RAG/GraphRAG pipelines and want someone who can ship to production — reach out.

• 📧 wleandro.oliveira@gmail.com
• 🌎 Open to remote (global, timezone UTC-3 / BRT)
• 💼 PJ contracts via Dewtech or full-time