AI-rewritten: This is a summary of an article from IEEE Spectrum, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

Gurmeet Saran is an engineer at Anthropic in San Francisco who works to ensure AI agents like Claude provide reliable and trustworthy answers. As a senior member of the IEEE, he develops infrastructure to understand how chatbots behave and verify that their outputs are accurate. His core responsibility involves ensuring data remains structured, correct, and private, noting that privacy must be built into the system’s architecture from the start rather than added later.

Saran previously worked at Yahoo, Intuit, and Meta, where he solved similar data quality issues. At Yahoo, his team created a central repository to define metrics like "daily active site users" so different teams could calculate them consistently. Later at Meta, he built Metric360 to fix errors in video view counts by establishing a single source of truth for all metrics across the company. He emphasizes that these problems were architectural, not just coding bugs, and that governance must be integrated into software design initially.

Now at Anthropic, Saran faces the challenge of making large language models auditable without exposing private user information. He compares his work to a doctor monitoring patient vital signs, warning that even slight miscalibrations in data can lead to serious errors in decision-making. He believes creating transparent and trustworthy AI systems is one of the most important engineering problems of the next decade. Recently joining IEEE in March 2025, he hopes the organization’s standards work will help shape the future of artificial intelligence infrastructure globally.