Alexandr Wang

Alexandr Wang grew up in Los Alamos, New Mexico, where his parents are physicists. He attended Los Alamos High School and competed in math, programming, and physics competitions during high school, also taking part in science quiz competitions and the school band. A student profile published in 2014 lists him as a finalist in the USA Computing Olympiad in 2012 and 2013 and a participant in the Mathematical Olympiad Summer Program in 2013.⁠[1][2]

On May 21, 2014, the American Association of Physics Teachers announced the 19 students selected for the U.S. Physics Team training camp, including Wang. The students were chosen from 4,277 first-round participants and went on to a training camp at the University of Maryland; the five-member traveling team for the International Physics Olympiad was selected separately.⁠[3]

After high school, Wang went to work in Silicon Valley, doing software engineering at Addepar and serving as a tech lead at QuoraQuoraQuora is an American question-and-answer and knowledge-sharing website and the company that operates it, Quora, Inc., whose core product centers on users asking, answering, and voting to rank content; since 2022, the company has also offered Poe, a chat platform that aggregates AI models from multiple companies.Open the full entry; his earlier biography also lists algorithm development work at Hudson River TradingHudson River TradingHudson River Trading is a proprietary trading firm headquartered at 3 World Trade Center in New York. Through its broker-dealer HRT Financial LP it trades for its own account and makes markets in US equities, options and futures, and it also handles part of the customer order flow of the retail broker Robinhood.Open the full entry. He later enrolled at the Massachusetts Institute of Technology to study artificial intelligence and took part in machine learning research, then dropped out to start a company.⁠[1][4]

Wang recalled in an interview that two of his Quora hackathon projects were incorporated into the actual product, which made him realize that small teams could quickly have broad impact. His research experience at MIT drew his attention to how hard it was for machine learning projects to obtain training data, which he saw as a starting point for a company.⁠[1]

In 2016, Wang co-founded Scale AI with Lucy GuoLucy GuoLucy Guo is an American technology entrepreneur and investor who cofounded Scale AI with Alexandr Wang and later founded creator platform Passes, where she is CEO. She also founded Backend Capital and has appeared in a School of Hard Knocks interview.Open the full entry and took part in the Y Combinator startup program. The company provides services around AI training data, combining machine learning tools with human annotation to process images, text, and other data for customers; autonomous driving was an early focus.⁠[5][1][6]

On July 16, 2017, Wang announced that Scale had closed a $4.5 million Series A round led by Accel, with Daniel Levine joining the board. At the time the company offered image annotation, optical character recognition transcription, categorization, comparison, and data collection through an API, and planned to use the funding to grow its team and product range.⁠[7]

On August 8, 2018, Scale announced an $18 million Series B round led by Index Ventures, with participation from Accel, Y Combinator, and others, and Mike Volpi joined the board. In the announcement, Wang likened data labeling infrastructure to cloud computing, saying it could lower the barrier to building machine learning applications; the company listed autonomous driving customers including GM Cruise, Lyft, Zoox, and nuTonomy.⁠[8]

Lucy Guo left Scale in 2018 after disagreements with Wang. She later told TIMETIMETIME is an American newsmagazine and media brand founded by Henry Luce and Briton Hadden, which published its first issue on March 3, 1923; it is headquartered in New York and known for its Person of the Year selection and the TIME100 list. Since 2018 it has been personally owned by Marc and Lynne Benioff; its current Chief Executive Officer is Jessica Sibley, and its Editor-in-Chief is Sam Jacobs.Open the full entry that one of their disagreements concerned whether contract workers were paid on time.⁠[9]

On August 5, 2019, Scale announced a $100 million Series C round led by Founders Fund, at a valuation of more than $1 billion. Its business had extended to natural language and content understanding, e-commerce search, mapping, robotics, and augmented reality, while its core process remained taking in raw data, labeling it with a combination of machine learning and human judgment, and returning training data to customers.⁠[6]

In August 2020, Scale launched Nucleus, which helps teams visualize and curate datasets, test and compare models, and share data. On December 1 of that year, the company announced a $155 million Series D round led by Tiger Global at a post-money valuation of more than $3.5 billion, along with its acquisition of Helia AI, which specialized in real-time video AI.⁠[10]

On April 13, 2021, Wang announced that Scale had closed a $325 million Series E round co-led by Dragoneer, Greenoaks CapitalGreenoaksGreenoaks is an American investment firm co-founded by Neil Mehta in 2012 and known for concentrated, long-term investing; it made an early, large bet on the Korean e-commerce company Coupang and has invested in companies including Wiz, Stripe, Databricks, and Anthropic. In May 2026, it co-led Anthropic’s $65 billion Series H funding round with Altimeter Capital, Dragoneer, and Sequoia Capital.Open the full entry, and Tiger Global; Jeff Wilke, the former CEO of Amazon’s worldwide consumer business, became an adviser to the CEO. The announcement showed the company expanding from autonomous driving data labeling into finance, logistics, e-commerce, and government projects, including document processing for Brex and Flexport.⁠[11]

On May 21, 2024, Scale announced a $1 billion Series F transaction, comprising new funding and a secondary sale, at a valuation of $13.8 billion, led by Accel. Wang summarized the company’s direction as expanding the data available, building data that pushes the frontier of model capabilities, and improving measurement and evaluation.⁠[12]

In the announcement he recalled that Scale had worked with OpenAI on early experiments in reinforcement learning from human feedback for GPT-2 and later extended the approach to InstructGPT; the company had also taken part in model red-teaming at DEFCON 31 and in large language model test and evaluation work for the U.S. Department of Defense.⁠[12]

On June 13, 2024, Wang published “Meritocracy at Scale,” proposing MEI, standing for merit, excellence, and intelligence, as the principle guiding hiring. He said he would interview every candidate or review their materials, argued for selecting people based on individual ability, skills, and work ethic, and maintained that merit-based hiring and diversity were not in conflict. These statements are his positions on the company’s talent policy.⁠[13]

On February 18, 2025, Theo VonTheo VonTheo Von is an American stand-up comedian and podcast host who presents the long-form video podcast This Past Weekend w/ Theo Von. A Louisiana native who first became known through MTV reality shows and comedy competitions, he launched the podcast in 2016; known for its free-flowing personal storytelling and guests from across the political spectrum, it ranked third among U.S. podcasts on Spotify in 2024, with guests including Donald Trump, JD Vance, Bernie Sanders and the chief executives of several technology companies.Open the full entry’s podcast This Past Weekend aired its 563rd episode featuring Wang, then the founder and CEO of Scale AI, discussing how artificial intelligence is changing the world, his decision to drop out of MIT, and whether AI is creating or eliminating jobs.⁠[14][14][14][14]

Asked whether his parents were proud of him, Wang said that when he first dropped out, it was not something one does in Asian culture; both his parents have PhDs, as do his two older brothers, and everyone in his family had gone all the way through school, so his parents were worried at first. He said he told them “a little bit of a white lie” that he would get the tech thing out of his system and then go back and finish school, but “obviously that hasn’t happened yet”; now, he said, they are very proud, and he owes everything to his parents, who taught him physics and math growing up, which is what let him do so well in competitions.⁠[15][15][15][15][15][15][15][15]

Wang said on the show that his parents are Chinese and his grandparents are from China. He said that for many Chinese Americans, Chinese culture is very different from the Chinese Communist Party and the current government: China is a culture and civilization that has existed for thousands of years, which is distinct from the current communist regime. He called what is happening to the Uyghurs horrible and said it was important to make sure there are governments that believe in democracy and liberty.⁠[15][15][15][15][15][15][15]

On March 7, 2025, Wang, Dan Hendrycks, and Eric Schmidt submitted “Superintelligence Strategy: Expert Version” to arXiv, with an updated version on April 14. The abstract proposes a three-part framework centered on deterrence, nonproliferation, and competitiveness to address the effects of superintelligence on national security and international competition; it is a policy framework proposed by the authors.⁠[16]

On May 1, 2025, Wang spoke with Gregory C. Allen at the Center for Strategic and International Studies (CSIS). He described data, compute, and algorithms as the three foundations of AI development, explained how Scale moved from autonomous driving to generative AI, model evaluation, and government applications, and set out his reasons for supporting the company’s work in U.S. defense.⁠[17]

In the discussion he proposed building a national AI data reserve, expanding energy supply, improving how government adopts AI, and strengthening talent education. He also stressed that military and government applications need mission-specific testing and continued monitoring for anomalous behavior after deployment. These are his policy recommendations and views on risk management.⁠[17]

In June 2025, Wang joined Meta as the company’s first chief AI officer, leading Meta Superintelligence Labs (MSL) and responsible for taking its AI strategy from vision to execution. On June 12, a Scale announcement confirmed that he would remain on the board after leaving the CEO role, and Jason Droege was named interim chief executive officer.⁠[18][19]

Meta invested $14.3 billion in Scale for a 49% stake. Scale’s announcement said the deal valued the company at more than $29 billion and stressed that it would remain independent; part of the investment proceeds was distributed to shareholders and vested equity holders.⁠[5][19]

On June 22, 2025, TIME reported that some Scale contractors had experienced delayed, reduced, or canceled payments, and cited Lucy Guo’s concerns about paying people on time. The report also included Scale’s response: the company said payment delays were extremely rare and that it was continually improving its processes.⁠[9]

On April 8, 2026, Meta released Muse Spark, MSL’s first model, for the Meta AI app and website. The company said the team had rebuilt its AI technology stack over the preceding nine months and that the model supported complex reasoning, multimodal understanding, and multi-agent collaboration; at launch, API access was a private preview for selected partners, and open-sourcing future models remained a plan.⁠[20]

On May 12, Meta updated the announcement to introduce more natural voice interaction, real-time vision, and shopping features, and began gradually expanding Muse Spark to its social apps and to Ray-Ban Meta and Oakley Meta glasses in the United States and Canada.⁠[20]

On June 12, 2026, WIRED reported, based on internal meeting records and employee interviews, that the Applied AI team formed to support MSL had run into discontent, with some of the roughly 6,500 engineers and product managers interviewed criticizing the way they were reassigned and the training data tasks. Meta did not comment on the report.⁠[21]

On June 15, WIRED reported that chief technology officer Andrew Bosworth acknowledged the reorganization had been poorly communicated and proposed improvements to management and career support. Applied AI head Maher Saba said employees could reapply for other positions within the company.⁠[22]

On July 24, 2026, Meta introduced new Meta AI features powered by Muse Spark 1.1, including making plans, connecting email and calendar apps, generating briefings and slides, and carrying out ongoing tasks. The company said the features would first roll out on the Meta AI app and website in some markets before expanding to other countries and to services such as WhatsApp.⁠[23]

On September 8, 2026, Meta released the personal agent Muse, and Wang pinned a post introducing the product on his X account. Built on Muse Spark, Muse can use a browser and connected apps to carry out tasks and was being rolled out gradually in the United States on iOS, Android, and the web.⁠[24][25]

Meta said Muse works in a separate virtual computing environment, that users decide which apps to connect and what permissions to grant, and that sensitive actions such as sending email or making purchases require user approval. A further confidential virtual machine plan in the announcement was slated to launch later in the year and cannot be treated as a capability available at launch.⁠[24]

On October 7, 2026, video journalist Cleo AbramCleo AbramCleo Abram is an American independent video journalist who hosts the YouTube science and technology explainer show Huge If True and the interview series Huge Conversations. A former video producer at Vox who worked on the Netflix show Explained and the YouTube Originals show Glad You Asked, she went independent in 2022 and quickly gained a large following with “optimistic” technology explainers, interviewing technology leaders such as Mark Zuckerberg, Jensen Huang and Sam Altman; by the end of 2025 she had more than 11 million subscribers across platforms.Open the full entry released an interview with Wang, titled “The Biggest Launch Since ChatGPT (Muse),” on her YouTubeYouTubeYouTube is Google’s video-sharing and creator platform, founded in 2005 by Steve Chen, Chad Hurley, and Jawed Karim and acquired by Google in 2006. Built around video uploading, viewing, and creator earnings, the platform has developed Shorts, YouTube Music, Premium, YouTube TV, and products for children, and is headquartered in San Bruno, California, United States.Open the full entry interview show HUGE* Conversations. It focused on what Muse can do, when it can be trusted, and how to make sure AI goes right, with chapters on how Meta handles and stores users’ information, how to teach an AI discretion, whether AI can police other AI, how to regulate AI, and what “personal superintelligence” means. The show notes that interviewees do not see questions ahead of time and have no editorial control.⁠[26][26][26][26][26][26][26]

Meta: Wang joined in June 2025 and is currently chief AI officer, leading Meta Superintelligence Labs. Official website: https://www.meta.com/ (opens in a new window).⁠[18]

Scale AI: The AI data company Wang co-founded with Lucy Guo, where he previously served as chief executive officer. Official website: https://scale.com/ (opens in a new window).⁠[5][19]

X: Public account: https://x.com/alexandr_wang (opens in a new window). Meta’s official profile page also links to his Facebook, Threads, and Instagram pages: https://www.facebook.com/alexandrwangg (opens in a new window), https://www.threads.com/@alexanddeer (opens in a new window), and https://www.instagram.com/alexanddeer/ (opens in a new window).⁠[25][18]

Photo: Taken by Village Global at The Grove on November 17, 2022, titled “Guillaume Pousaz & Alexandr Wang”; this is the cropped and rotated version of Wang from Wikimedia Commons. Original: https://www.flickr.com/photos/villageglobal/52531276663/ (opens in a new window). License: CC BY 2.0, https://creativecommons.org/licenses/by/2.0/ (opens in a new window).⁠[27][28]