Train Guard
Team consisting of SNCF Data Scientist/NLP Engineer and Data Engineer, plus an Analytics Engineer, expert in Python, LLMs, dbt, Airflow, and Streamlit apps.
Loom Video
Project Description
Taking incidents report (PDF) to categorized information of relationships of the root causes with GraphRAG (Neo4j), this project utilizes the llm-graph-builder-mcp architecture to perform detailed safety analysis. We convert the unstructured SNCF 2023 Annual Safety Report into a structured knowledge graph, defining nodes for accidents, root causes, regulatory bodies, and mitigation plans. The system links fatal accidents (e.g., Level Crossing and Worksite incidents) to their corresponding intervention programs (:ActionPlan), allowing stakeholders to query complex relationships like the effectiveness of safety initiatives ([:TARGETED_BY]) against core performance indicators. This provides superior querying and visualization of safety data compared to traditional tabular formats.
Prior Work
none, new to all these tools
