Itamar Friedman

Itamar Friedman is the co-founder and chief executive of Qodo, which builds AI-powered code review, testing, and governance for engineering teams. His premise: generating software faster creates little lasting value unless organizations can verify its quality, preserve architectural understanding, and decide when human judgment remains essential.

From machine vision to code integrity

Friedman studied electrical engineering at the Technion–Israel Institute of Technology and worked at Mellanox before co-founding Visualead, where he served as chief technology officer and developed computer-vision technology for visual QR codes. In 2017, Alibaba moved to acquire Visualead technology and bring members of its engineering team into a Tel Aviv research operation. Friedman subsequently led machine-vision work at Alibaba, including neural-architecture search, model training, and automated machine-learning tools.

In 2022, he and Dedy Kredo founded CodiumAI to help developers understand, test, and review code. The company expanded from automated testing into code integrity across software development and became Qodo in 2024. A $40 million Series A brought its funding to $50 million at the time; Friedman subsequently announced a $70 million Series B in 2026.

• Flow engineering over prompt engineering. Friedman co-authored AlphaCodium with Tal Ridnik and Kredo, developing a test-driven, multistage approach that iteratively improves generated code. Reliability comes from the surrounding workflow: generating tests, evaluating candidate solutions, and incorporating feedback.

• Independent code review. Writing code and evaluating it demand different judgments. Qodo’s open-source PR-Agent reflects Friedman’s focus on specialized reviewers that detect consequential defects while considering security, architecture, maintainability, and team standards without inundating developers with irrelevant warnings.

• Organizational context as infrastructure. Effective review depends on information beyond a codebase: architectural decisions, service contracts, incident investigations, previous pull requests, and engineering conventions. Friedman’s context engine organizes that knowledge so review decisions remain intelligible to humans and usable by coding agents.

• Adversarial collaboration between coding agents. Friedman uses command-line workflows to connect specialized coding, testing, and review systems, while distinguishing simple pipelines from genuine agent interoperability. His practical escalation rule requires the coding agent and independent reviewer to resolve disagreements; when they cannot, a human decides.

• Software-graph governance. Friedman wants review systems to evaluate relationships among repositories, services, contracts, and simultaneous changes—not merely individual pull-request diffs. Teams can gradually automate approval and blocking as their standards become explicit and review results earn trust. He calls the translation of accumulated engineering experience into auditable, actionable judgment artificial wisdom.