AILY LABS
Data Scientist (Madrid). Shipped LangChain agents to production, with Langfuse for tracing—retrieval, tool calls, and guardrails.
NLP and time-series forecasting in SQL/Python. Owned Airflow and Docker pipelines on AWS; cut an internal ETL from ~1.5h to ~15s.
Graph RAG for shop-floor manuals
Built Graph RAG over factory equipment manuals for manufacturing lines. Ingest ran OCR on scanned PDFs, then unstructured.io to keep structure—not flat text, but tables, figures, and diagrams, each tagged with page and section coordinates.
Indexed into Neo4j as an equipment graph (machines, components, faults, procedures). LangChain retrieves against the graph; a production API returns cited passages with exact page and section so other models and apps can call it with sources.
On the shop floor: when something fails, the operator sees what broke and where to look in the manual—the right page and section, not a vague summary. Same class of shop-floor intelligence Aily ships at enterprise scale; see their Sanofi case study (manufacturing agents across hundreds of lines).
Graph RAG pipeline
Scanned equipment manuals from the factory—PDF and OCR on diagram pages.
Operator gets the fault, the manual location, and the exact page—not a vague summary.