CTO & Co-Founder @ Sokrates.AI — building AI-powered education systems

Generative AI · Multi-Agent Systems · RAG · Web Accessibility · Edge Computing & 5G

LinkedIn  •  Academic page  •  Email

I'm a Computer Engineering student at UFRJ (Escola Politécnica) and CTO & Co-Founder of Sokrates.AI, an award-winning startup applying Socratic AI to education. I work end-to-end on modern AI products — from multi-agent architectures and RAG pipelines to cloud infrastructure and accessible interfaces — backed by a research foundation in edge computing and 5G networks.

• 🚀 Leading tech & architecture at Sokrates.AI — multi-agent tutor (LangGraph, RAG, multi-provider LLM gateway) on AWS + GCP + on-prem, used by 69 teachers across 19 institutions; in a pilot with RNP's Network College, 30 of 52 students joined voluntarily and 72.7% found the mentoring useful
• 🔬 Former research intern at GTA/UFRJ — built an Android toolkit that collected ~20k latency samples on real 4G/5G networks; two published papers + public dataset
• ♿ Web accessibility (WCAG 2.2 / eMAG / ABNT NBR 17225) — audit of UFRJ's accessibility portal with 17 proposed fixes
• 🦀 Currently leveling up in Rust for high-performance backends

• 🥇 1st place — IEEE AI Pioneer Summit 2025 (Brazil) with Sokrates.AI
• 🌎 Best Startup of Latin America — IEEE Region 9 Entrepreneurship 2025 (16 startups, 6 countries)
• 🌍 Finalist (4 of 38 startups) — IEEE Technology Summit 2026, AI Founder Showcase (Budapest)
• 🎓 Honorable mention — SIAC/UFRJ undergraduate research conference

• Vehicle Mobility Impact on Performance of Multi-Access Edge Computing — VehiClouds'24 (IEEE CloudNet), international
• Caracterização da Latência de Borda sob Efeito de Mobilidade a Partir de Dados Reais — WGRS'24 (SBC), national

• grounded-rag — Multi-agent RAG with a groundedness critic: a LangGraph pipeline (Retriever → Answerer → Critic) that verifies every answer against sources and retries retrieval when evidence is missing. Runs 100% locally (Ollama + Chroma).
• a11ybar — Zero-dependency accessibility toolbar mapped to WCAG 2.2 / eMAG 3.1 / ABNT NBR 17225, with opt-in VLibras — born from my accessibility-audit work at UFRJ. Live demo.
• emag-audit — CLI auditor for the Brazilian accessibility standards: 18 checks mapped to WCAG 2.2 + eMAG 3.1 that output instructional Markdown reports (diagnosis → fix proposals). Field-validated against WAVE/AMAWeb findings.
• WolfLatency — Android client/server toolkit (Kotlin/Compose) that measures edge (MEC) latency under mobility on real 4G/5G networks: ~20k samples collected in Rio de Janeiro, public dataset, data behind both papers above.
• Edge Latency under Mobility — Python analysis (pandas · scikit-learn · matplotlib) turning those measurements into the figures and statistics reported in the publications.

Focus areas: multi-agent AI & RAG · generative-AI applications · accessible web platforms · mobile (Flutter/Kotlin) · cloud infrastructure

• 💼 LinkedIn: luizfecristino
• 🌐 Academic page: gta.ufrj.br/~cristino
• 📧 Email: lfelipe@poli.ufrj.br