Introduction
After the bankruptcy, Suresh Mehta found duplicate supplier names in three old spreadsheets. One had an accent mark, one had an abbreviation, one had a wrong country code. None looked dramatic. Together they had helped make bad decisions look reasonable.
Suresh lives in Zurich and works as a data quality manager for AI-supported purchasing. His interest in data began with the expensive consequences of bad master records.
Story of the Path into AI
His import business failed under pressure from supply-chain disruptions, currency risks and decisions based on poor information. Looking back, Suresh saw old prices, double entries, unclear product codes and missing fields. The mess had not belonged to any one person, which was part of the problem.
He began with spreadsheets rather than large models. He learned data cleaning, classification and quality metrics. His first tool checked supplier data for duplicates, missing fields and contradictory labels. It immediately made useful finds and one embarrassing mistake: it merged two suppliers with similar names but different roles.
Suresh added a review step and began treating data quality as a process, not a one-time clean-up.
Current Work
Today Suresh prepares data for AI-supported purchasing decisions in a trading company. In one pilot he stopped a forecast because product numbers from two countries had been wrongly combined. The correction was dull, slow and essential.
Purchasing models now produce more stable results because the foundation is cleaner and easier to trace. Suresh does not mind that his work is rarely visible in presentations. Bad data can make an impressive model dangerous. Good data often looks like nothing happened.
Personal Advice
“Before you admire the AI, check the table it stands on,” Suresh says. He advises companies to respect the unglamorous work of names, codes, dates and definitions.
Key Facts
Age and place: 48, Zurich.
Background: bankruptcy, international supply chains, pragmatic restart.
Entry into AI: tool for supplier-data duplicates and contradictions.
Focus today: trade and data quality.
Typical tools: data cleaning, classification, quality metrics.
Werkstattnotiz
Suresh’s favourite dashboard is a plain list of corrected supplier records. It has no dramatic chart. He keeps it because it prevented a false forecast. His current worry is that teams call data “clean” once it looks tidy, not once it has been understood.