The National Association of Software and Service Companies (NASSCOM) K-Tech Centre of Excellence for Data Science and Artificial Intelligence (AI) has launched a machine learning operations (MLOps) playbook. It is a compendium of MLOps implementation frameworks and industry best practices. MLOps is a set of practices for data scientists and operations professionals that increase the quality, simplify the management process, and automate the deployment of ML models.
The compendium was launched in association with a private strategic partner and incorporates extensive market research outcomes that have been supported by an accounting and research firm from the industry. The playbook will function as a blueprint for the deployment of MLOps in companies.
A report from the government’s AI portal stated that organisations across industries are spending on AI/ML initiatives to generate business insights from data but are not able to scale their ML models to production at an enterprise level. Manual, time-consuming efforts are required for ML model monitoring, ML model retraining with new data, and subsequent deployment to production environments. ML experiments are not reproducible, and data scientists do not have access to technology infrastructure that can auto-adapt to their needs.
According to a report on the top ten data and analytics technology trends for 2020, 75% of companies will push their AI competitive frontiers by 2024. They will move away from pilot projects to scale AI adoption in production. Organisations that are better prepared to channel their AI investments into the appropriate tools, technologies, practices, frameworks, and skillsets will be better equipped to reap five-fold benefits in their ROI. Thus, organisations will need to solve the challenges that arise due to poor data infrastructure, which is where NASSCOM expects its MLOps playbook to be most helpful.
The playbook was launched at the latest edition of AI Parley, which is an initiative powered by NASSCOM that focuses on creating learning facets for the AI community and where thought leaders can share opinions and experiences on contemporary themes. A panel discussion was held after the launch of the playbook. The discussion explored the issues regarding ML adoption at scale, the need for MLOps to overcome the challenges, and key drivers of MLOps adoption that fuel the success of AI investments in organisations.
Like industry players, the Indian government is leveraging the potential of ML in its operations. Earlier this year, the country launched a project to speed up the delivery of several public services using AI/ML. The project, the Digital Government Mission, will break down silos and make government systems more intelligent. It will help citizens learn about schemes that they are eligible for, and government benefits they can receive. The idea is that the benefits should be offered to citizens without them having to apply for them. For instance, a student who qualifies for a government-funded scholarship scheme will get an automatic alert from the concerned department, instead of the student having to enquire and apply for the same.
OpenGov Asia quoted an official saying that the concept of citizens receiving services without asking for them is necessary, and with AI and ML, the predictive part becomes easier. People should be able to get automatic renewals every time their driving licence or passports expire since all the data is already available with the government, the official explained.