Client-facing B2B data analytics for the Government of India, a top-3 global payment network, and a Big-4 audit firm: translating high-stakes business problems into data-driven models and dashboards.
Tap any item below to read the full story.
Government of India: Operation Clean Money
High-Risk Account Detection: surfaced ~$3B in high-risk accounts from $175B in deposits; $130M+ seized, contributing toward $1.5B recovered.
Operation Clean Money was a national-priority project: analyze $175B in post-demonetization cash deposits to identify tax evaders, under intense political scrutiny. The core data-science team's complex ML model was taking months, and government stakeholders were losing patience.
I ran a parallel 80/20 track. Instead of a full model, I defined simple heuristic and outlier rules: government employees with massive cash deposits, accounts depositing 100x their declared income, petrol pumps three standard deviations above their segment norm. Within a week we isolated a 1,000-account "sore thumb" shortlist (~$3B in suspicious deposits) from millions of flagged accounts, and presented it as a "Phase 1" strike list.
It gave the client an immediate win: raids on ~900 of those accounts seized $130M+ in assets ($90M in cash), and it bought the data-science team the time to complete the full model that ultimately identified $1.5B.
- Response-Likelihood Targeting: Cut on-ground investigation effort by 30% with a model that prioritized flagged accounts by their likelihood of responding to a tax notice.
Enterprise Analytics
Conflict-of-Interest Prototype (Big-4 auditor): built a no-budget interactive demo in Excel/VBA; saved ~2 dev-months (~$60K).
On a high-stakes engagement, the data-science team's conflict-of-interest detection algorithm was stalling, and the client, with no UI to interact with, was losing confidence. With no UI developers, budget, or time, I built a fully functional prototype in Excel + VBA: an input dashboard that simulated the back-end logic on a sample dataset and visualized the detected conflicts and relationships.
The client could finally "touch and feel" the solution and gave invaluable feedback that fed the final product's requirements, and it eliminated the need for a separate web prototype, saving an estimated two man-months (~$60K) of development.
- Drove revenue for a top-3 global payment network by enabling 20% more purchase transactions, and a 36% lift in payment-switching via target-list optimization.
- Conducted due diligence on a USD $200M energy-sector investment for a PE firm.