2
2
Brazil
1 year of experience
Entusiasta de tecnologia, inovação e inteligência artificial. Gosto de transformar ideias em projetos reais, unindo criatividade e propósito. Sou movido por desafios e acredito que cada hackathon é uma oportunidade de aprender, colaborar e criar algo que realmente faça diferença

QUBIX is a decentralized GPU cloud platform designed to solve the global shortage of high-performance hardware used for AI training, inference, and scientific computing. Today, demand for GPUs far exceeds global supply, causing long waiting lists, high prices, and limited access on centralized clouds like AWS, Google Cloud, and Azure. QUBIX transforms this broken model by connecting users directly to independent GPU providers worldwide through a trustless, transparent, blockchain-powered marketplace. Using the Qubic blockchain, QUBIX enables instant, feeless payments, secure wallet-based identity, and verifiable transaction integrity. Providers earn QUBIC for renting their hardware, while users gain immediate, on-demand access to compute at up to 80% lower cost. QUBIX supports multiple environments—Jupyter, SSH, VS Code, APIs—and automatically manages rental sessions, costs, monitoring, and expiration. The result is a platform that is scalable, permissionless, and community-driven. Anyone can supply GPUs. Anyone can consume them. No intermediaries. No barriers. QUBIX democratizes access to compute and builds the foundation for an open, decentralized AI infrastructure.
7 Dec 2025

The **AI Tech Recruiter** is an agentic AI solution that transforms technical recruitment by automating the end-to-end hiring funnel. Built on **IBM watsonx Orchestrate**, the agent operates as a reasoning engine, interpreting complex job requirements in natural language (e.g., "Find a Senior Python Developer with FastAPI in Brazil"). **Agent Workflow:** 1. **Intelligent Discovery:** The agent invokes a **Custom Skill: Candidate Search** to proactively query active profiles on GitHub, ranking them based on repository quality and stack fit (using the IBM Granite 3.1 8B model). 2. **Engagement:** Upon human approval, the agent sends personalized outreach messages via channels like Telegram. 3. **Agentic Intelligence (MCP):** The agent enters a "Waiting Stage" (managed by the **Model Context Protocol - MCP**), freeing up resources. When the candidate replies, an n8n webhook triggers the callback, and the workflow is **autonomously resumed** from the pause point. 4. **Scheduling:** The agent uses **Native Skills** for Google Calendar and Gmail to check the manager's availability, handle time zone differences, and send a confirmed interview invitation. **Impact:** The project demonstrated a **96\% reduction** in direct hiring logistics costs and a **200\% improvement** in candidate response rates, allowing recruiters to focus solely on strategic decision-making.
23 Nov 2025