In October 2018, in Palo Alto, a quiet problem stopped being quiet. Machine learning models were already making real decisions with real consequences, and no one in the room could fully explain why. Krishna Gade had lived that tension firsthand building explainability systems for Facebook News Feed. Amit Paka had shipped global products at Samsung where scale turns small errors into loud ones. Manoj Cheenath knew how fragile complex systems become once they leave the lab. Fiddler AI was born to stare straight into the black box and not blink.
Fast forward to late January 2026 and that conviction just cleared another gate. Fiddler AI closed a $30 million Series C led by RPS Ventures, with Timothy Murphy backing the bet alongside returning believers from Lightspeed VP, Lux Capital, Insight Partners, Capgemini Ventures, Dallas VC, Dentsu Ventures, and Mozilla Ventures. New strategic muscle came in from LG Technology Ventures, Benhamou Global Ventures, and LDV Partners. The post money valuation lands at $235 million, pushing total funding to $100 million and putting real weight behind a very specific thesis.
The thesis is simple to say and brutal to execute. Autonomous agents do not fail politely. They fail across chains of models, tools, APIs, and policies, often faster than humans can notice. Fiddler AI positions itself as the neutral control plane above that chaos, delivering observability, evaluation, policy enforcement, and auditable governance for compound AI systems. Not a plugin. Not a point fix. A system of record for what agents do, why they did it, and how to intervene in under 100 milliseconds when things drift sideways.
Enterprises are responding. Revenue is up more than four times in the last eighteen months. Fortune 100 and Fortune 500 teams are already in production. Nielsen Chief Executive Officer Karthik Rao has publicly endorsed the platform. The US Navy cut model update time by 97 percent. Brex expanded monitoring across business units. Healthcare, finance, insurance, and government teams are adopting Fiddler because regulators do not accept vibes as evidence.
This round is about scale and timing. Agentic workflows are leaving experimentation and walking directly into regulated, high-risk environments. Fiddler’s batteries included Trust Models, deep explainability roots, and framework agnostic design meet enterprises where they actually live. Krishna Gade, Amit Paka, and Manoj Cheenath are not selling magic. They are selling visibility, control, and accountability at the moment those words stop being optional.
