NITI Scope is an independent analysis of India's public procurement - roughly 3.9 million tenders and 3.17 million unique contract awards, 2011–2026, Central and state. It exists to answer one question anyone should be able to ask of their government: when it goes shopping with public money, does anyone actually compete to sell to it?
Procurement is where policy meets money. Competition is the mechanism that is supposed to protect that money - and whether competition actually happened is measurable, if the data is public. India publishes more of that data than most countries realise. NITI Scope turns it from an unreadable firehose into something a journalist, researcher or citizen can actually interrogate: competition rates, integrity red-flags, spending patterns, vendor concentration and corporate networks, sliced by ministry, state and sector.
It is not an accusation engine. A single bid is a risk signal, not proof of wrongdoing. The goal is to show where the soft spots are and let others look closer - not to name culprits.
Millions of raw procurement records → a classification and analysis pipeline on Google Cloud (an analysis VM, a key-gated JSON API on Cloud Run, and this static site) → aggregates the browser can render without touching a database. Corporate networks are reconstructed by matching winning vendors to the government's company registry (MCA) to surface shared addresses, emails and co-bidding.
Every figure is labelled robust or directional. Contract values are treated as directional because buyers publish them inconsistently; bid-count and date figures are the robust ones. The methodology is public, and so is a running corrections log.
NITI Scope was built using agentic AI - the same applied-AI practice its author works in. The human contribution is the judgement: what to measure, what to trust, what to retire, and the rigour of checking it. In that sense the project is a working demonstration of what agentic development can do in the right hands, as much as it is an analysis of procurement.
The methodology is easier to show than to assert. Three decisions that shaped NITI Scope:
After publishing, I found ~30% of the source's award records were exact duplicates from overlapping scrapes. I deduplicated, restated every count in public, and showed the headline rates barely moved - because ratios survive that kind of fault.
I had flagged a "March rush". A reader pushed back; I tested it by weighting on money instead of counting contracts, found it did not hold, and demoted it from red flag to calendar effect - on the record.
A buyer showing zero contract values turned out to explain 97% of India's entire missing-value gap - one company. I quantified the scale from its own accounts and drafted the RTI to recover the rest.
NITI Scope is built and maintained by Nivedita Pandey, founder of Dendrons.ai (agentic AI). Her work sits where applied AI meets public systems: a Prime Minister's Fellow on Digital Public Infrastructure and an advisor to the Reserve Bank Innovation Hub, she has also built Asia's first digital-gold fintech product and Microsoft's PE/VC partnerships vertical across JAPAC, and trains civil servants and PSU executives in applied AI. She holds an MBA from The Wharton School and a degree in Computer Science Engineering.
On this project she framed the questions, chose what to measure and what to trust, ran the audits that corrected the site's own figures, did the investigative work (the corporate networks, the BHEL disclosure finding), verified the international benchmarks against official sources, and drove it end-to-end - from raw data to a published site and API - using agentic development throughout.
What it demonstrates: data analysis and integrity at scale, cloud data engineering, statistical reasoning, product judgement, investigative research and clear writing - held together by a bias toward honest, checkable claims and a working command of agentic AI.
Open to senior roles in applied AI, digital public infrastructure, product and strategy - across Singapore, the UK and the US. Happy to walk through how any of this was built.
Data: CPPP & state procurement portals, via the open scrape at tender.sarthaksidhant.com. Analysis & curation: Dendrons.ai. Aggregates licensed CC BY-NC 4.0.