Designed end-to-end GenAI architecture to anticipate self-consumption (autoconsommation) contract eligibility and sharpen EDF’s commercial positioning.
Architecture
- RAG Ingestion Layer: Smart meter readings, consumption profiles and tariff grids vectorised
- LLM Scoring Engine: Mistral generating eligibility scores and contract-fit rationale from retrieved context
- Real-Time Decision API: FastAPI/OpenShift consumed by commercial tools via REST
- Explainability Layer: Surfacing key drivers per prospect to sales teams
Impact
Reduced contract analysis cycle from days to minutes, enabling proactive targeting of high-value AutoConso prospects.
Stack
Python, LangChain, Mistral, FastAPI, Oracle DB, OpenShift, pg-vector