i have asthama so I built a policy tester for Delhi’s air with AI agents
My project is gcorp , and this article explains that , it tests an air policy and simulates the city in a 3d enviornment along with the agents working to show the consequences of it
There will be thousands of people participating in this , thousands using ai to make their project like i did as well , but what stands out is how you think about a problem and the bottleneck is up there in the brain not the coding , its how you communicate and present your idea , you may have the wierdest thing put together but u need to sell it properly with ur pitch
so thats what i did here , i made gcorp .
Delhi's air is quite popular , but for the wrong reasons , and i myself face consequences of that on daily basis , but its not just the air , its the weekly or i would say quite often there is a change in a policy or a new system is created to counter that , and thats being done by the higher authorities ,
but they dont test it ,they debate among themselves and just put it on us and we have to just be ready for what comes our ways , so i created a tool where we can simulate and test it first and then if its a good enough impact , we proceed with it
The tool works with a team of ai agents , so ill introduce you with the team
The team
- The Compiler turns the planner’s sentence into a clear proposal.
- The Source analyst asks whether the policy targets enough of the pollution to matter.
- The Engineer estimates how much the emissions actually change.
- The Epidemiologist works out who breathes the air, and for how long.
- The Economist asks who pays for it.
- The Externalities analyst checks what drifts in from outside the city’s border, like stubble smoke.
- The Skeptic hunts for the weakest assumption.
- The Simulator and the Verifier aren’t AI at all. The simulator gives the range, and the verifier strikes any claim without a source.
- The Brief writer turns what survived into the final note for the planner.
-
Now Aws is a big part of how this was made possible , from deploying on agentcore runtime , using gateway as the way for tools and policies for the agent ruleset , dynamo db for database , sns for aqi alerts in notifications over mail , cloudwatch for system health check , bedrock for ai models , and lamda for the api's when called
Over the 3 days i made this for the bharat builds hackathon , i learnt about how systems on such scale break so often if you dont pay attention to them , and this tool helped me understand a lot of new ways to make ai agents work together specially on aws
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article