Ali-Reza Adl-Tabatabai is a computer scientist and co-founder of Gitar, the code-validation company acquired by Sonar in May 2026. His career spans compiler research, reliability engineering, and developer platforms. With Gitar, he addresses a growing constraint on software delivery: AI can accelerate code production while leaving engineers with more changes to review, test, and repair.
From optimized code to developer infrastructure
Adl-Tabatabai earned his doctorate in computer science at Carnegie Mellon University in 1996, researching source-level debugging of globally optimized code. Compiler optimizations can move operations and eliminate variables, making a running program difficult to reconcile with its source. His research developed techniques for tracking those transformations so a debugger could present useful source-level information without restricting optimization or instrumenting the program. When a transformation could not be hidden, the debugger would expose its consequences to the programmer.
He subsequently served as a director and senior principal engineer at Intel Labs, leading research into programming-language technologies and hardware support. By 2012, he was a member of Facebook’s HipHop compiler team, which worked to improve PHP performance through compilation and a dedicated virtual machine. His later engineering leadership at Google included Site Reliability Engineering across products, cloud systems, and technical infrastructure.
At Uber, his focus expanded to the tools used by an entire engineering organization. He led the Developer Platform organization, responsible for frameworks and tools for building, deploying, and managing software across the company. In 2023, he co-authored Uber’s account of its generative-AI experiments, covering prototypes for coding, test generation, code quality, and reducing operational work. The work explored AI across the development lifecycle, with safe adoption treated as part of the engineering problem.
Experience with Uber’s build systems, code analysis, testing platforms, and automation helped shape Gitar’s founding vision: a modern, end-to-end development platform. The increasing volume of AI-generated changes sharpened its focus on validation. Gitar emerged from stealth in April 2026 with $9 million in seed funding led by Venrock, with participation from Sierra Ventures.
Making validation an active workflow
Adl-Tabatabai connects developer productivity with the work required to trust a change. In his account of the thinking behind Gitar, fast feedback, useful signals, and less manual coordination are central to keeping engineers productive. His public argument for agentic validation starts with the pressure AI puts on CI and code review: more and larger pull requests can force teams to slow delivery or approve changes with less scrutiny.
Several ideas define his response:
Connected review and repair:Gitar connects code review, failure diagnosis, fixes, and testing. Its agent can identify an issue, commit a proposed repair, monitor continuous integration, and iterate when checks fail. Distinguishing code failures from flaky tests and infrastructure noise helps engineers act on the right signal. The purpose is to reduce repeated handoffs between a review comment, a developer, and a failing pipeline.
Intent-aware review: Adl-Tabatabai and Gautam Korlam distinguish passing checks from fulfilling the purpose of a change. A successful build establishes that the configured checks passed; it does not establish that the implementation delivers the requested behavior. Their vision for Gitar at Sonar brings codebase context, team conventions, and the intent behind a change into validation alongside conventional quality checks.
Program analysis combined with agents: Their integration strategy pairs SonarQube’s repeatable program analysis with Gitar’s contextual reasoning and ability to propose repairs. Static-analysis findings can inform AI review, while generated fixes can pass through deterministic gates before landing. This describes a direction for deeper integration, without treating automated review as a guarantee of correctness.
Automation earned through trust: Adl-Tabatabai describes teams building confidence step by step, extending automation from review and CI diagnosis to fixes, approval, and merging under rules they set. Human supervision and team-defined gates let organizations decide how much responsibility to delegate as they gain experience with the system.
Ali-Reza Adl-Tabatabai explains how Gitar automates the path from code review and CI failures to a green pull request—and why workflow orchestration, gradual trust and program analysis matter as much as the agent.
AI-generated PR volume moves work into CI and review, where failures can add hours or days of coordination and delay.
Agentic validation connects review and CI diagnosis to retries, repairs and a loop toward a green PR. Team-defined conditions govern automatic approval and merge.
A validation-specific harness lets Gitar optimize token cost, precision and coverage, while orchestration and model routing remain separate responsibilities.
Combining agents with SonarQube program analysis is the proposed next step toward better precision and coverage; the talk presents the direction without a comparative evaluation.
Scaling agents means changing how teams specify work, measure results and grant automation permission. Sonar’s Tariq Shaukat and Gitar’s Ali-Reza Adl-Tabatabai explain the organizational changes—and the gradual path from advisory code reviews to automatic merges.
Scaling agents requires changes to workflows, tooling and people’s roles; a successful pilot does not automatically spread its working habits.
▶ Watch ↗