Barakaldo, Spain

For me, technology is not just a tool — it's an endless rabbit hole. I pull one thread, new branches appear, and sooner or later they connect. That's what keeps me building.

I learned to make decisions and deliver value leading teams in retail. As a software developer at CEESA, I brought that standard to a real ERP (Vespera Solutions, Velneo): real-time stock, tax and traceability, plus AI for inventory anomaly detection, order forecasting and KPI analysis, a Telegram bot (real-time business alerts + RAG assistant over the ERP with Python + FAISS), and agent-automated documentation.

I'm a DAM graduate (Tartanga, 2024–2026), now specializing in AI & Big Data. Certified in the Google × Kaggle 5-Day AI Agents intensive (prototype to production, multi-agent, ADK, evaluation, security) and the Anthropic Claude API + Claude Code 101 tracks. I build with Java, Python, Velneo, PHP · PostgreSQL, MySQL, MongoDB · scikit-learn, pandas, NumPy, XGBoost · PyTorch, TensorFlow/Keras, OpenCV · RAG, local LLMs, MCP · Git, Docker, Linux, with SDD/TDD discipline: I explore with AI, agents and data science, but I ship on engineering fundamentals.

• liveInterviewRAG — AI voice interviewer: digital twin answering with my real profile. STT → RAG → LLM → TTS pipeline with phrase-by-phrase SSE streaming and a reactive 3D avatar.
• fraud-detector — Hybrid fraud detection in 3 layers: deterministic rules + XGBoost + context. SHAP explains every score; a local LLM drafts the reports without deciding.
• Proyecto-Velneo — Full ERP build (stock, tax, real-time traceability) with AI integrations, developed at CEESA.

• Email: mikelromerohomobono@gmail.com
• GitHub: https://github.com/mikelrh-dev
• Location: Barakaldo, Spain · Spanish (native) / English (B2) / Euskera (B2) · Available for hire