Software Engineer building AI-powered products — React · Next.js · TypeScript on the front, Node.js · FastAPI · Python on the back. 6+ years shipping across SaaS, healthcare, e-commerce, Web3 and real-time analytics, with a broader tech background since 2010.
Lately I go deep on production LLM patterns — retrieval (RAG), agentic workflows and multi-model orchestration — and self-host most of it on a Proxmox homelab.
Based in Brazil.
Frontend: React · Next.js · TypeScript · Vue/Nuxt · Tailwind CSS Backend: Node.js · FastAPI · Python · tRPC · Prisma · PostgreSQL AI / LLM: LangChain · multi-provider LLMs · RAG / retrieval · conversational memory · multimodal · agentic workflows Web3: Solidity · Hardhat · ethers.js · ERC-721 Infra: Proxmox homelab · Docker · Ansible (IaC) · Vercel · CI/CD
- tradeOS — real-time market-context dashboard: 51 assets × 5 timeframes with a decision engine (trend, liquidity zones, setup readiness) over live Binance data.
- Mr. Knowledge — live-audio desktop copilot: a Rust capture pipeline, real-time transcription and streaming LLM suggestions in a low-latency overlay.
- Signals — self-hosted signal-intelligence engine: ingests Telegram (MTProto) → PostgreSQL, with a Claude vision pipeline that reads price levels straight from chart screenshots.
- Books — turns EPUB/Markdown into structured notes with an LLM and serves them in a Next.js editorial reader, auto-deployed on the homelab.
- Homelab — Proxmox + Docker + Ansible (IaC) with a custom Next.js / React Flow ops dashboard: live topology, service health, GitOps auto-deploy and a Telegram AI agent.
→ 20+ more, with write-ups, at leonidasamaro.com