At Razorpay, I built an automated GMV monitoring pipeline that compares each in-store merchant against 28-day baselines, ranks alerts by severity, creates tickets for owners, and posts daily summaries to Slack.
I also built a Python pipeline that classifies POS support tickets with an LLM, and onboarded two product domains to Razorpay’s AI self-serve analytics platform with a 17-metric catalogue.
My payments analysis uncovered ₹1.5 Cr in late-authorised payments that banks auto-refunded across 950+ merchants. I independently rebuilt the numbers to challenge a bank’s ₹1.8 Cr claim, and flagged anomalies including a UPI routing misconfiguration behind 90% of monthly QR failures.
For a platform migration, I built a seven-stage Airflow pipeline and multi-bank Tableau dashboards, pre-validating 15,000+ merchants and catching 190K+ record discrepancies before go-live. Earlier, at Polco Maven, I built web-scraping and NLP insight reports using Python, NLTK, and SpaCy.
