Joel Pobar is a technologist and engineering leader known for his work on software platforms and artificial intelligence projects. He has held senior roles at major technology companies, including leading infrastructure and engineering at Anthropic and serving as a senior engineering organizational leader at Meta, where he worked on AI initiatives including PyTorch and was associated with the Superintelligence team. ^[1]

Education

Joel Pobal completed a Bachelor of Information Technology with Honours in Computer Science at the Queensland University of Technology (QUT) between 1999 and 2001. ^{[3] [4]}

Career

Joel Pobar began his professional career in Brisbane as a Systems and Network Administrator at Quicknet Internet Provider, where he was responsible for maintaining IT infrastructure across multiple platforms, including Windows, Linux, FreeBSD, and Cisco-based systems, between 1999 and 2003. Concurrently, he worked as a Research Assistant at the Queensland University of Technology (QUT), participating in projects related to programming languages, distributed computing, and software adaptation frameworks. In 2002, he completed an internship at Sun Microsystems in Menlo Park, contributing to the development of a code partitioning system for an internal Verilog compiler targeting ASIC-based execution environments.

In 2003, Pobar joined Microsoft as a Program Manager on the Common Language Runtime (CLR) team. His responsibilities involved work on several components of the .NET framework, including reflection, code generation, and delegate mechanisms, as well as contributions to the Shared Source CLI (SSCLI). He later collaborated with the Microsoft Solutions Development Centre (SDC) as a contractor, participating in technical architecture and engineering efforts across multiple large-scale software initiatives. In 2010, he served as a contracted engineer at Macquarie Bank in Sydney, where he helped develop trading infrastructure designed for high-throughput and low-latency operations.

Pobar moved to Facebook (now Meta) in 2012, where he held successive engineering leadership roles for over a decade. Early in his tenure, he managed teams responsible for core software infrastructure projects such as HHVM (a just-in-time compiler for PHP), the Hack programming language, and the Flow static type checker for JavaScript. Between 2015 and 2019, he directed initiatives focused on performance infrastructure across Facebook’s applications, overseeing optimization efforts for mobile and backend systems. From 2019 to 2023, he worked with AI and machine learning infrastructure teams, contributing to internal platforms used in model deployment, orchestration, and large-scale training workflows, including work involving PyTorch and FBLearner.

In 2023, Pobar joined Anthropic, where he is part of the technical staff focused on engineering infrastructure for inference systems supporting large language models. In 2024, he became the first Venture Partner at TEN13, a venture capital firm with investment interests in artificial intelligence. His role involves advising on technical aspects of AI-related opportunities and contributing to the firm’s investment strategy.

Pobar has also been involved in technical publishing and conference presentations. His contributions include articles for MSDN Magazine and talks at industry events such as the GOTO Conference in Aarhus, where he spoke on the use of functional programming techniques in applied computing contexts. ^{[2] [1] [3] [4] [5]}

Superintelligence Team

In June 2025, Meta Platforms established the Meta Superintelligence Labs (MSL), a division focused on research and development related to artificial general intelligence (AGI). The unit was created with the objective of consolidating Meta’s AI efforts under a unified organizational structure.

The lab is jointly led by Alexandr Wang, formerly CEO of Scale AI and appointed as Meta’s Chief AI Officer, and Nat Friedman, former CEO of GitHub, who co-leads product and applied research initiatives. Concurrent with the formation of MSL, Meta acquired a 49% stake in Scale AI for approximately $14.3 billion. This acquisition aligned with broader efforts to expand data infrastructure and increase recruitment capacity in the AI sector.

Following the creation of MSL, Meta initiated a targeted hiring strategy to attract AI professionals with experience at organizations such as OpenAI, Google DeepMind, Anthropic, and Apple. Reports indicated that compensation packages offered during this process included substantial bonuses. Recruits included Daniel Gross (formerly of Safe Superintelligence), Ruoming Pang (previously with Apple's foundation model group), as well as Trapit Bansal, Ji Lin, Shuchao Bi, Hongyu Ren, Jack Rae, and Pei Sun, among others. These individuals brought expertise in areas such as multimodal models and reasoning systems.

Joel Pobar was among those who joined the MSL team. Prior to this, he worked at Anthropic in a role centered on large-scale inference infrastructure and had previously spent over ten years in engineering leadership roles at Meta.

The launch of MSL followed internal evaluation of Meta’s existing model development efforts, including the open-source Llama 4. It also took place in the context of increasing competition within the AI field, particularly from companies such as OpenAI, Google, Anthropic, and DeepSeek. MSL’s stated focus includes research into general-purpose AI systems, with applications under consideration such as automated content generation, personalized tools, and extended-reality interfaces. ^{[6] [7]}