An Artificial Intelligence Life Cycle: From Conception to Production

Abstract: Drawing on our experience of more than a decade of AI in academic research, technology development, industry engagement, postgraduate teaching, doctoral supervision and organisational consultancy, we present the 'CDAC AI Life Cycle', a comprehensive life cycle for the design, development and deployment of AI systems and solutions. It consists of three phases, Design, Develop and Deploy, and 17 constituent stages across the three phases from conception to production of any AI initiative. The 'Design' phase highlights the importance of contextualising a problem description by reviewing public domain and service-based literature on state-of-the-art AI applications, algorithms, pre-trained models and equally importantly ethics guidelines and frameworks, which then informs the data, or Big Data, acquisition and preparation. The 'Develop' phase is technique-oriented, as it transforms data and algorithms into AI models that are benchmarked, evaluated and explained. The 'Deploy' phase evaluates computational performance, which then apprises pipelines for model operationalisation, culminating in the hyperautomation of a process or system as a complete AI solution, that is continuously monitored and evaluated to inform the next iteration of the life cycle. An ontological mapping of AI algorithms to applications, followed by an organisational context for the AI life cycle are further contributions of this article.

• The paper introduces a comprehensive framework that delineates 17 stages across Design, Develop, and Deploy phases for AI systems.
• It outlines rigorous data preparation, ethical reviews, and explainability methods to refine AI model development and evaluation.
• The framework promotes operational integration and strategic alignment by leveraging MLOps and hyperautomation for scalable AI deployment.

Overview of the CDAC AI Life Cycle

The paper "An Artificial Intelligence Life Cycle: From Conception to Production" (2108.13861) introduces the CDAC AI Life Cycle, a comprehensive framework for the design, development, and deployment of AI systems. The framework consists of three core phases—Design, Develop, and Deploy—and encompasses 17 distinct stages. This life cycle aims to provide researchers and engineers with a structured pathway from the conception to the operationalization of AI solutions, offering guidance based on extensive experience in AI research, application development, and consultancy.

Design Phase

The Design phase emphasizes the contextual problem formulation and the importance of situating AI projects within established ethical guidelines and frameworks. This phase involves:

• Problem Identification and Formulation: Defining problems through environmental, entity, and data perspectives.
• Literature and Ethics Review: Assessing state-of-the-art AI models, pre-existing solutions, and ethics guidelines.
• Data Preparation: Establishing a unified data infrastructure with attention to data ethics and governance.

The Design phase sets a foundation for aligning AI initiatives with both operational goals and ethical standards, crucial for the legitimacy and sustainability of AI applications.

Develop Phase

The Develop phase transforms raw data into functional AI models and consists of:

• Data Exploration and Augmentation: Engaging in activities such as feature engineering and addressing class imbalances.
• AI Model Construction: Building initial models, executing parameter selection, and leveraging transfer learning techniques.
• Model Evaluation and XAI: Applying metrics to assess model performance and adopting explainability tools like LIME and SHAP to interpret results.

The Develop phase is central to refining data into valuable insights, offering an iterative loop for improving model performance and establishing objective benchmarks.

Deploy Phase

The Deploy phase concerns the operationalization of AI solutions, focusing on:

• Computational Performance Evaluation: Analyzing computational metrics crucial for sustainable deployment.
• Model Deployment and Serving: Selecting appropriate approaches for real-time or batch execution.
• Operational Integration: Utilizing MLOps/AIOps methodologies for systematic model updates and management.
• Hyperautomation: Integrating AI capabilities into automated systems to enhance efficiency and innovation.

The Deploy phase ensures that AI models not only function optimally under computational constraints but also seamlessly integrate into existing workflows, promoting efficiency and scalability.

Contributions to AI and Strategy

The paper includes contributions beyond the life cycle stages, such as:

• Ontological Mapping of AI Algorithms: Defining a structured mapping of AI capabilities—Prediction, Classification, Association, Optimization—to applications, facilitating collaboration between technical experts and stakeholders.
• Organizational Strategy: Situating the AI Life Cycle within broader organizational contexts, linking technological functions to strategic decision-making processes.

These contributions aim to bridge technical developments in AI with high-level organizational strategy, ensuring that AI initiatives support long-term goals and operational coherence.

Conclusion

The CDAC AI Life Cycle offers an exhaustive, structured approach to the creation and operationalization of AI systems. It delineates clear stages that not only foster technical excellence but also promote ethical integrity and strategic alignment. By providing a detailed roadmap, this framework aids researchers and practitioners in systematically advancing from the conception to the deployment of impactful AI solutions. The paper anticipates that the life cycle will stimulate further discussion, knowledge dissemination, and policy formulation, contributing to a transparent and informed AI ecosystem.