Unlocking Efficiency with Generative AI: A Case Study

Welcome to our blog where we dive into the world of generative artificial intelligence (AI) and its transformative impact on businesses. In this post, we are excited to share a guest collaboration with Vicente Cruz Mínguez, Head of Data and Advanced Analytics at Cepsa Química, and Marcos Fernández Díaz, Senior Data Scientist at Keepler, where we explore how generative AI is reshaping the energy sector.

Generative AI is revolutionizing the way organizations operate by leveraging the power of machine learning algorithms to streamline processes, generate human-like content, and drive innovation. At Cepsa Química, a leading player in the chemical manufacturing industry, and Keepler, a cloud-centered data services consulting company, our collaboration has resulted in the implementation of a generative AI assistant to enhance the efficiency of the product stewardship team.

The Impact of Generative AI

Cepsa Química, a pioneer in the production of linear alkylbenzene and phenol, is committed to sustainability and decarbonization in line with Cepsa’s Positive Motion strategy for 2030. As part of this journey, the company’s Digital, IT, Transformation & Operational Excellence (DITEX) department has been working on democratizing the use of AI to drive value creation.

Within the Safety, Sustainability & Energy Transition team at Cepsa Química, product stewardship plays a critical role in ensuring regulatory compliance for the marketed products. The team manages a vast collection of compliance documents, a task that previously consumed a significant amount of time. By leveraging generative AI techniques, we have been able to streamline the compliance query process and accelerate response times.

The Solution

Our approach to product stewardship with generative AI involves utilizing Large Language Models (LLMs) trained on extensive internet data to provide dynamic and adaptable responses to compliance queries. Through a Retrieval Augmented Generation (RAG) approach, we offer up-to-date information without the need for continuous model retraining, making it ideal for the fast-paced world of regulatory changes.

The solution we built comprises four key functional blocks:

• Input processing
• Embeddings generation
• LLM chain service
• User interface

By dividing the solution into two modules – one for batch processing input documents and another for answering user queries through inference – we have created a seamless and efficient system.

Challenges and Solutions

During the development process, we encountered challenges related to data preprocessing and evaluation of model responses. To tackle these challenges, we implemented strategies such as chunking data for context, selecting models with large context windows, and generating query variants for better retrieval accuracy.

Furthermore, we established an automatic evaluation system that benchmarks responses against ground truth data, enabling us to refine and improve the system continuously.

Achievements and Next Steps

Through the implementation of our generative AI assistant, the product stewardship team at Cepsa Química has seen significant improvements in query times, answer quality, and operational efficiency. Building on this success, we are now looking to expand the use of generative AI across other business areas within the company.

Conclusion

Generative AI is revolutionizing the way organizations operate, and our partnership with Cepsa Química and Keepler exemplifies the transformative power of this technology. By leveraging Amazon Bedrock and RAG techniques, we have been able to streamline compliance processes and drive efficiency across the organization.

If you’re looking to explore the world of generative AI for your business, connect with our team at Generative AI on AWS or explore PartyRock for quick application development.

About the Authors

Vicente Cruz Mínguez is the Head of Data & Advanced Analytics at Cepsa Química, with expertise in big data and machine learning projects across various industries.

Marcos Fernández Díaz is a Senior Data Scientist at Keepler, specializing in end-to-end machine learning solutions.

Guillermo Menéndez Corral is the Sr. Manager, Solutions Architecture at AWS for Energy and Utilities, driving innovation through cloud technology in the energy industry.