The context
Container flows involve heterogeneous data and planning requirements: availability, capacity and appointments. The goal was to make this data usable for logistics stakeholders.
My role
I led the development of AI solutions: a GenAI insight engine for querying CO₂ KPIs, OCR + LLM document pipelines and time-series forecasting modules.
From data to use
I connected RAG pipelines and vector stores, forecasting models and FastAPI services. Docker, Kubernetes, Airflow and Scaleway support the integration and deployment of these services within the platform.
Results and publications
The AI solutions developed contributed to a 10% improvement in operational efficiency in logistics flow management. Two KES 2024 publications present the work on container availability forecasting and port digital twin modelling.
What this case demonstrates
Translating business needs into AI pipelines and services, and describing methods in publications. Production code and client data are not public.