Ankit Jain is the co-founder and chief executive of Aviator, a developer-productivity company building tools to verify AI-generated code. He is tackling a fundamental problem for software teams: how to maintain engineering quality, architectural judgment, and shared understanding when coding agents produce changes faster than humans can review them.
Jain worked as an engineer at Adobe and Google before holding engineering leadership roles at Homejoy, Shippo, and Sunshine. He was also an entrepreneur in residence at Unshackled. He co-founded Aviator in 2021, joining Y Combinator’s Summer 2021 batch. The company initially focused on merge automation, reliable builds, and deployment workflows—bringing Google-style engineering productivity to teams without Google’s internal infrastructure. Jain argues that productivity comes from improving the systems surrounding development, testing, review, and deployment, not simply measuring how busy engineers appear.
He also leads The Hangar DX community, a network of senior engineers and DevOps practitioners, and Xoogler, the former-Google alumni network. The Hangar DX Podcast explores engineering productivity and developer experience.
Rethinking how software earns trust
- Implementation intent belongs in the review. Fixed specifications miss decisions that emerge through issue tickets, product requirements, and conversations with coding agents. Jain proposes preserving those interactions and converting them into acceptance criteria, giving reviewers access to what engineers intended, what changed during implementation, and which architectural choices require scrutiny.
- An AI slop registry makes engineering judgment reusable. Repeated review comments can become explicit guardrails against unsafe patterns, broken conventions, and violated invariants. Combined with acceptance criteria, these accumulated rules generate executable test plans that improve as teams capture more recurring mistakes.
- Deterministic verification should anchor automated trust. Aviator Verify runs test plans against live previews and supplies behavioral evidence, including screenshots and database-state checks. Jain favors deterministic checks wherever practical and reserves model judgment for situations that require it, shifting human attention toward intent, observed behavior, and architectural decisions.
Jain increasingly views these verification systems as essential platform engineering: software companies using coding agents must build the execution environments, feedback loops, and guardrails that make machine-generated development dependable. His vision for AI-native engineering organizations places people in charge of product intent, technical standards, and the infrastructure through which agents write and verify software.
