Ali Farhadi

Overview

Ali Farhadi (born 1982) is a computer scientist and a professor in the Paul G. Allen School of Computer Science & Engineering at the University of Washington^[1]. From 2023 to 2026 he was chief executive of the Allen Institute for AI (Ai2)^[2]. His career has passed through three of the settings in which AI research is done: a university department, a large technology company, and an independent research institute.

Education

Farhadi received his PhD in computer science from the University of Illinois in 2011, with David Forsyth as his adviser^[3]. The doctoral adviser relationship is the ordinary way a researcher learns the craft of the field — how a problem is chosen, how a result is argued, how work is written up — and it is the reason advisers are usually named alongside a researcher's early career.

Recognition

In 2017 he won a Sloan Research Fellowship^[1]. Fellowships of this kind are given to researchers near the beginning of their independent careers, and they are one of the standard markers by which a young faculty member's work is judged from outside their own institution.

Industry and the institute

Before his appointment at the institute, Farhadi most recently led Apple's machine learning initiatives^[3]. That placement is notable because it puts him among the researchers who moved from academic posts into senior roles at large technology companies during the period when machine learning became central to consumer products.

On 20 June 2023 the Allen Institute for AI announced that he would become its chief executive, describing him as rejoining the institute as its second CEO; he took office on 31 July 2023^[2][3]. The word "rejoining" is the operative one: the appointment returned him to an organisation he had worked at earlier in his career, rather than bringing in an outsider. He was the second person to hold the CEO position in the institute's history^[3].

Open releases and departure

During his tenure, Ai2 went through fully open releases of its OLMo, Molmo and Tülu model families^[2]. "Fully open" is a stronger claim than "open weights": it normally means that the materials needed to reproduce the work — not only the trained model but the data and code around it — are published as well, so that outside researchers can inspect and rebuild what was done rather than only use the finished result.

On 10 March 2026 the institute announced that Farhadi was stepping down, after about two and a half years in the role^[2].

Where he sits in the field

Three things about this career are worth holding together. First, it spans the places where AI is made: a public university, a company that ships products at global scale, and a research institute funded to publish its results. Second, his time at Ai2 coincided with a deliberate choice by the institute to release models openly, which lets other groups check the work and build on it. Third, the move back and forth between these settings is itself common in AI — people carry methods and habits of work with them, and the boundaries between academic and industrial research are porous in practice.

What the record shows is a scientist who held senior positions in all three settings and, at the institute, was responsible for a period in which openly released models were the organisation's visible output^[1][2][3].