Artificial Intelligence at population scale
How machine learning and analytics were deployed to address nationwide COVID-19 impact and recovery in Telangana, India.
Nivruti Rai, Country Head for Intel India and VP of the Data Center Group at Intel, took the stage at the recent AWS Partner Summit to walk us through an exceptional feat. A team of multidisciplinary experts that stood on the shoulders of AWS technology to build a lifesaving app at population scale in record time.
As the unfortunate chain of events unfolded in India, everyone from citizens to government to enterprise were severely affected. The lack of understanding of the novel coronavirus coupled with the sheer density of populace made the challenge insurmountably complex and widespread. The rapidly changing situation gave the challenge another dimension making it overwhelming and disorienting at each step. With the GDP of the nation reflecting a fifth of the world, the damage was far too devastating and crippling. It is at a time like this that technologies that were slow to adopt and evolve became center stage highlighting the potential.
Intel India, along with AWS India, Fractal AI, and a group of 17 tech-service groups came together to take on the daunting dilemma. This challenge of epic proportions could only be solved by the collaboration of leaders, academics, and technologists – with technology serving as the backbone of the solution. The conglomerate started working on an open Pandemic Platform that relied upon open architecture, cloud-based access, and multiple city and citizen centric usage – all done at population scale.
leading the pandemic response
The key aspects of the solution resonated with the capacities that AWS brought to the table. For a nation of 1.2 billion people, the solution had to be fully scalable and needed to be built and function in real-time. The collaborative efforts and ethical nature of the service ensured that it was open to any user nationwide that could help. Finally, the solution had to be modular and flexible so that the rapidly adapting insights could continue throughout the most critical periods of time.
Every step of the data pipeline towards a decision was supported by an array of AWS technology and services evolving through the core workflow from Ingest, Prepare, Analyze to finally Act. AWS fueled the development of the valuable platform with databases and data feeds ready within minutes, data wrangling and automated ETL channels, ML environments and querying capabilities, and dashboard and summary analysis systems. Beyond these functionalities, AWS also maintained the liabilities – privacy, security, and compliance. Removing the heavy lifting of infrastructure management and security allowed the team to focus on the logic of the application while maintaining a brisk pace of development and deployment. In the words of the Nivruti, the “rich and prolific ecosystem of AWS India” made it clear that the scope of these technologies would deliver more than the solution was originally built for.
from idea to solution
The Proof of concept (PoC) of the solution was built within days and a dashboard of the results was presented to multi-level task forces from government and administration to health diagnostics and control to security and mobility to citizen education. The two states that tested the PoC were split into their respective district zones with each having its own model to manage and predict the spread of the virus. As the critical problems to solve for kept evolving, a five-part categorization appeared: Trace, Contain, Predict, Minimize, Unlock to help citizens prepare for the pandemic and guide them towards eventual normalcy.
As the platform gathered more and more data, ML possibilities only grew. From understanding the spread patterns of the virus to understanding the disease itself, so experts began to develop the platform further. With patient medical history and treatment plans added into the mix, the dimensions became increasingly complex and sensitive. AI/ML solutions that are easy to build, convenient to deploy, and safe to access proved to be the revolutionary solution towards this crisis. These tools helped discover patterns and dependencies, and the insights produced helped protect the vulnerable. On the non-technical side, quick-to-access dashboards provided leadership teams clarity of the problem and transparency of the solution.
All this was done at a hyper-local level with tools supplied on an open, accessible and performant platform, all in real-time in the race from data to decision.
