Darius Gros

AI products and custom software for SMEs and mid-market companies Designed and built a complete RAG pipeline (production MVP): FastAPI backend, PostgreSQL with pgvector, React frontend, LangGraph orchestration with hexagonal architecture. • Developed modular AI components: document segmentation, semantic enrichment, classification, eligibility, evaluation, and structured generation modules. • Deployed document recognition agents with iterative review loops; planned roadmap includes fine-tuning,observability, and LLM cost/latency optimization. • Delivered Next.js 15 / TypeScript / Tailwind v4 web projects for external clients, including motion design withFramer Motion.

Real-time event augmentation product, sold B2B to production agencies and brands. • Designed and built a real-time backend (FastAPI, WebSocket, PostgreSQL) deployed on a sovereign EU VPS; zero US cloud dependency, fully GDPR-compliant. • Built an automated NLP filtering pipeline (HuggingFace, Detoxify, French sentiment models) for live audience message moderation at events of 100 to 5,000 participants. • Developed the participant web app as a Next.js PWA accessible via QR code — no install, all smartphones supported. • Electron Desktop Application for light technician (back end to Electron via SSE), light CC web frontend for users via QR code

• Industrialized Spark / Delta Lake pipelines with end-to-end CI/CD (GitHub Actions, Airflow, Databricks). • Translated business needs into data solutions, coordinating across two business directions and the central data platform team. • Led legacy stack migrations, drove cloud cost optimization, and mentored junior data engineers. Deployment time reduced from 2h to 17 min (−85%); ~20% savings on cloud infrastructure.

• Built ETL pipelines for product analytics and marketing; developed time-series forecasting models and ran A/B tests. • Delivered decision-support dashboards (Tableau, Power BI) used by business stakeholders.