Enabling Enterprise-Wide Automation & AI Implementation for a Billion Dollar Pharma

5 min

IN A NUTSHELL

Machine analyzed 400,000+ audit findings

2500 broad topics produced

Utilized core data science libraries to adhere to org principles

About Client

One of the largest pharmaceutical companies headquartered in the UK, the client has research and IT centers around the globe to help in their innovation drug discovery, manufacturing and supply.

EMPLOYEES

83,000

LOCATION

United Kingdom

INDUSTRY

Life Sciences

01

What was the challenge

The client’s audit analytics relied on manually assigned event categories, prone to errors that undermined leadership decisions. The system's inability to assign multiple categories or quantify labeling certainty limited audit accuracy, highlighting the need for a precise, automated solution.

02

How we solved the problem

Ciklum joined the client as a trusted data science and AI partner to transform their audit analytics. Tasked with solving challenges in event categorization, Ciklum built a powerful Machine Learning pipeline that analyzed over 400,000 audit events. By leveraging unsupervised learning techniques, the solution accurately assigns context-driven category tags, ensuring precise insights. This innovation empowered the company to confidently address critical decisions, whether improving lab training or refining drug prescriptions, fostering smarter, data-driven outcomes

Custom AI Solution Development

Ciklum leveraged its expertise in custom product solutions to develop a sophisticated AI system tailored to improve decision-making for business stakeholders. Our solution prioritized transparency and traceability, addressing limitations in explainability commonly found in LLMs like ChatGPT while eliminating risks such as hallucinations.

Clustering-Based Approach

Recognizing the need for semantic grouping, Ciklum identified this as a clustering challenge. Our team explored supervised and unsupervised machine learning techniques, focusing on topic modeling and classification. This ensured event categorization was rooted in meaningful, data-driven insights.

Explainability Through Coherence Modeling

By linking machine-defined topics back to the original data using coherence modeling, Ciklum delivered clear and reliable evaluation metrics. This approach empowered stakeholders with explainable insights, essential for confident decision-making.

Continuous Improvement Pipeline

Ciklum implemented a production-ready AI pipeline designed for continuous improvement. This strategy not only enabled real-time inference but also ensured the model’s capabilities evolved to meet future demands.

Enhanced Stakeholder Reporting

The team integrated machine-learned labels into the existing data corpus, enabling seamless visualization in stakeholder reports. This alignment ensured the AI insights were both actionable and easily accessible.

Rapid Business Confidence Building

Within months, Ciklum demonstrated its ability to build business confidence. By mastering the complexities of the existing process, mapping a feasible journey, and delivering a bespoke technical platform, the team ensured a smooth and impactful implementation.

Improved User Experience

Ciklum’s focus on user-centric design ensured the AI model delivered an exceptional stakeholder experience. Clear, explainable results and actionable insights solidified the solution’s value in enhancing decision-making processes.