A leading natural resources company that engages in exploration, development, and processing of petroleum, iron ore, copper, and coal wanted to replenish its resource base and enhance its portfolio by leveraging technology and innovation. The goal was to expand its exploration program for petroleum and copper.
The client initiated a joint global endowment study to explore future growth opportunities and yet-to-find (YTF) volume using data analytics and artificial intelligence.
Key Challenges
- Inconsistent implementation models: Lack of common coding language, libraries, real-time collaboration tools, inefficient data process between data management and data scientists along with poor model reporting capabilities lead to various inconsistencies in the ML processes which resulted in difficulty to measure the accuracy (number of correct predictions on a dataset) of the model leading to lack of trust on the model and underlying algorithm used.
- Focus on technology rather than business: Emphasis on technical aspects of configuring compute resources and cloud environments rather than solving real business problems. Data scientists had limited exposure to cloud and compute technology and faced a steep learning curve diverting their focus to technology rather than business.
- Difficulty to operationalize: Lack of process in building data science models and development. Absence of coding guidelines and centralization led to challenges in operationalizing even though the proof of concept (POC) was successful.
Ready to experience?
TALK TO EXPERTS
The Solution
Infosys developed a platform-centric approach to support, integrate, and enable automation of the data science pipeline.
- Modular platform-centric approach: Infosys leveraged the core concepts and design approach of the Infosys Enterprise Data and Analytics Platform to design and build a client-specific, state-of-the-art enterprise data management platform with data modeling techniques supported by robust governance and operations.
- Accelerated cloud factory capability: Infosys engineers contributed to client’s cloud factory based on the experience from our flagship cloud service Infosys Cobalt. This helped create re-usable assets for the future and accelerated the time to market not just in the program but across the client’s enterprise.
- Phased approach: A phased approach was adopted for this project. The MVP was delivered on-premises and a stable production-ready release was delivered in hybrid mode. The last phase was the movement to “ever-green and cloud” where the platform is constantly upgraded to provide data scientists with a state-of-the-art, optimized, data science platform for efficiency and cost optimization.
Cost-efficient, scalable, and extendable solution on cloud
- Set up a continuously operating platform with constant upgrades for reduced cost and increased efficiency
- Co-created templates and blueprints for re-use by collaborating with client’s nascent cloud-factory
- Established standards for platform delivery and provided operational guidance for future platforms in the client’s portfolio
- Optimized the cloud usage with resource and pricing recommendations which resulted in saving more than 30% of cloud cost from go-live to steady state
Benefits
Designed a modern, standardized, scalable, cloud-hosted, collaborative data science platform that abstracts the complexities of the underlying computation technology and enables data scientists to solve business problems (such as finding petroleum and copper resources)
Achieved 20-50% effort savings on data modeling by abstracting the technical complexity and computing platform
Built a pre-configured AutoML pipeline for faster ROI with the ability to move the prediction model from design to production in less than 10 minutes
Increased data science efficiency with a cloud-hosted integrated data science platform that provides AutoML capabilities
Democratized data science, paving the way for citizen data scientists to make predictions and drive decisions with a few clicks and no code
