Cristian I.

I build production AI systems, not thin API wrappers. Whether it's a focused MVP or a full-scale platform, I deliver complete pipelines from data ingestion through retrieval and LLM orchestration to deployed, documented APIs. 15+ years in data-intensive software. The last 3+ years focused entirely on LLM-powered products across regulated and high-stakes domains: legal compliance, HR, energy markets, healthcare, financial data, and CX analytics. Before AI engineering, I spent 8 years as a scientific software engineer at CNRS (French National Research Centre), building research-grade crystallographic software and managing HPC infrastructure. That background in scientific computing and peer-reviewed research is what separates my work from typical AI integrations: I build systems with rigour, not just speed. What I deliver: • RAG pipelines over large document corpora with semantic search, hybrid retrieval, and citation tracking (Pinecone, pgvector, sqlite-vec) • LLM orchestration — multi-step autonomous workflows, streaming chat, prompt management, structured output with validation • Production FastAPI backends with auth, caching, background processing, and OpenAPI docs • Data ingestion and extraction pipelines (PDF, Excel, XML, OCR) with AI + traditional fallback strategies • Full deployment: Docker, Nginx, Gunicorn, DigitalOcean, Azure Current production systems include: — Multi-country HR compliance audit platform with branded report exports — Portuguese labour law compliance engine with OCR and vector-indexed legal corpus — AI sales outreach platform orchestrating Perplexity research + Claude composition — CX analytics platform with streaming LLM chat, survey RAG, and embeddable sessions — Real-time lightning risk monitoring system processing EUMETSAT satellite data — Automated payroll extraction from mixed-format archives with banking field validation I own the full backend scope with minimal oversight. Direct communication, clear progress, blockers flagged early.