Hiding in plain sight: Can AI help uncover tender secrets?

I'm here to talk about an open-source project proposal (Here's the prototype: https://tenderai.vercel.app/). We know thousands of crores of Rupees are "spent" in India every year, on thousands of "tendered" projects. They're published in newspapers everyday, with oddly specific amounts for oddly specific tasks: Cleaning of a 750m road stretch for Rs 1,76,476.54. How do these numbers come up? What happens to these projects? Who's selected? What's the required qualification to bid? How many such projects are cancelled, re-tendered? What's the cost escalation? How many of these miss original deadlines?

How many crores of public money is spent in my area?

We need an open-source, open-data, citizen-led effort to create a project that tracks this publicly-available but hard to track – buried deep inside layers of random, non-functional websites and captchas – data into a streamlined, filterable, longitudinal record of public spending.

There's no better disinfectant than sunlight, and its time we start using AI to shine the light on India's dismal public works.

• How can we use agents to orchestrate the collection of hard to find, terribly buried government data

• Using AI workflow frameworks like Langchain/Mastra to read and parse PDFs etc

• Can we create a collaborative project to make this happen? Better urban areas are what's holding India back. We need to fight.