Student Team
Tanay Goel
TenderShield Project
An AI layer that makes corruption visible, documentable, and preventable before public procurement money is lost.
India’s public procurement funds hospitals, roads, and schools. When irregularities occur, the loss is not just financial: it can mean a classroom, road, or hospital that never reaches citizens.
Platforms like GeM and CPPP moved procurement online, but they digitised the process, not the judgement. TenderShield adds an intelligence layer on top of public records.
Each stage leaves a paper trail. TenderShield is designed to read that trail at scale and surface risk signals before money is lost.
Requirements tailored to favour a specific vendor.
Collusion signals that suppress genuine competition.
No clear documented scoring rationale.
Inflated invoices, ghost work, and execution gaps.
Documents exist, but no one connects them at scale.
Flags suspicious specification language designed to favour a single vendor.
Detects collusion signals such as identical pricing, rotation patterns, and suspicious withdrawals.
Aggregates compliance, litigation, and performance history into a single risk indicator.
Creates an immutable record of every evaluation decision: who scored what, and why.
Prioritises and routes complaints with supporting evidence, reducing noise while protecting sources.
Uses public procurement information only. It is not surveillance; it is intelligence on public data.
One state, such as Uttar Pradesh or Maharashtra, one department, and one tender category: road construction.
Six months: enough time to observe a full tender cycle, short enough to act on findings.
Pre-agreed metrics, independent review, and a clear go/no-go decision at the end.
Reduction in tenders with only one bidder, a key indicator of spec rigging.
Increase in unique vendors bidding, a signal of genuine competition.
Reduction in cycle time, since delays can be a cover for manipulation.
Savings against estimated project cost, the most direct measure of value.