David Luan

On May 25, 2011, TechCrunchTechCrunchTechCrunch is an American technology news outlet founded in 2005 by Michael Arrington and Keith Teare and headquartered in San Francisco; it mainly covers technology companies, startups, and venture capital, and hosts events such as TechCrunch Disrupt. Its ownership passed successively through AOL, Verizon, and Yahoo, and in 2025 Regent announced its acquisition.Open the full entry published the list of the first Thiel Fellowship class, Peter ThielPeter ThielPeter Thiel is an American entrepreneur and investor, a co-founder of PayPal and Palantir Technologies, and the first outside investor in Facebook. He is a partner at the venture capital firm Founders Fund and founded the Thiel Fellowship, which funds young people to leave college and build new things; his public skepticism of higher education, his support for Trump in 2016, and his funding of litigation against Gawker Media have drawn controversy.Open the full entry’s “20 Under 20” Thiel Fellows; he was listed as David (Jiageng) Luan, the only fellow in the robotics group.⁠[1][2]

Digital Trends’ profile of the fellows the same day wrote that a lifelong passion for robots drove his wish to reinvent the consumer robotics industry, with the aim of putting robots into homes to do chores and tasks of various types and skill levels.⁠[2]

Years later he said on the Latent Space podcast: “I’m personally fascinated by robotics. I’ve always loved robotics.”⁠[3]

Yale Scientific Magazine wrote in April 2012 that Darren Zhu, a member of the same first Thiel Fellowship class, moved to Palo Alto, CaliforniaCaliforniaCalifornia is a U.S. state on the Pacific coast; its capital is Sacramento and its largest city is Los Angeles. California was admitted to the Union on September 9, 1850, as the 31st state. Its population at the 2020 Census (April 1, 2020) was 39,538,223, the largest of any state.Open the full entry, to start a company, and that he teamed up with Luan, another Yale UniversityYale UniversityYale University is a research university in New Haven, Connecticut. Its predecessor, the Collegiate School, received a charter in 1701. Trustees chose New Haven in 1716; the institution became Yale College in 1718 and Yale University in 1887. Yale provides undergraduate, graduate and professional education and has 14 undergraduate residential colleges. As of October 2026, its president was Maurie McInnis, who took office in July 2024. In 2026, Yale expanded aid for incoming undergraduates and celebrated its 325th anniversary.Open the full entry student and Thiel Fellow, to found Dextro. The article describes Dextro as a biotech robotics company that aimed to automate large-scale experimental procedures in order to reduce tedium and error in scientific research.⁠[4][1]

The article states that the two later abandoned the project: the high cost of the system’s infrastructure limited the market to large biotech companies, and many potential customers planned to cut their research and development spending.⁠[4]

AlleyWatch wrote in 2017 that Dextro was founded in 2013 by Sanchit Arora and Luan, who served as co-founder and CEO; TASER International’s February 2017 press release, for its part, describes Dextro’s history up to the acquisition as “the two years since their founding.”⁠[5][6]

Xconomy writes that the idea took shape while he was living in San Francisco, where he saw more and more cameras being built into robots and connected devices. He and Sanchit Arora had both studied computer vision for robotics; the two first built a simple API that let connected devices do “smart things” with what they saw, and in his words “hacked together a primitive version” to see what people wanted. Developers and large companies then approached them about using it for social-media analytics and content categorization, which led to a more robust effort. “We spent a lot of time building our technology,” he said.⁠[7]

In a 2015 AlleyWatch interview he said that Dextro was started when they realized that perception, a robot’s ability to make sense of complex human scenes through its cameras, was one of the largest problems facing widespread robotics adoption; after early prototypes drew attention, they found the system could generalize to video-understanding problems in many markets. He said half of Dextro’s team had a robotics background, from MIT, iRobot, the GRASP Lab, SRI and elsewhere; the company was based in SoHo, New York City, and chose New York because it could assemble a team of machine-learning PhDs and master’s graduates there.⁠[8]

AlleyWatch introduced him in 2015 as the “Founder and CEO of Dextro.” He said Dextro made video “searchable, discoverable, and understandable by machines,” and that its “Sight Sound and Motion” platform analyzed the visuals, audio and motion in a video rather than relying on text tags; the company had completed its first financing with New York investors and was growing both its engineering and business teams.⁠[8]

Xconomy writes that Dextro launched a tool for discovering publicly shared Periscope live streams, which at the time covered only Periscope content; the company had raised $1.7 million from angel investors including Esther Dyson and the co-founders of Medidata and Skype, from the venture firms RRE Ventures and Two Sigma Ventures, and from Yale University.⁠[7]

AlleyWatch wrote in 2017 that Dextro’s seed round totaled $1.71 million from YEI Innovation Fund, Connecticut Innovations, Two Sigma Ventures, RRE Ventures and KBS Ventures, with individual investors including Esther Dyson, Jaan Tallinn and Semil Shah; the article calls Dextro the first deep-learning company to analyze livestreams in real time.⁠[5]

On February 9, 2017, TASER International announced two acquisitions to create an artificial-intelligence division, Axon AI; one of them was Dextro, the other a computer-vision team that Fossil Group had obtained through its acquisition of Misfit. The press release states that Dextro makes videos discoverable and searchable using machine learning and calls its system the first computer-vision and deep-learning system to make the visual contents of video searchable in real time; the two acquisitions brought a team of nearly 20 researchers and engineers. The release refers to him as “Dextro co-founder and CEO.”⁠[6]

In the release he said: “Dextro has built a highly advanced video analytics technology driven by machine learning,” and that in combination with Axon “we create an incredibly powerful AI solution tailored to the unique requirements of public safety.”⁠[6]

In its interview after the acquisition, AlleyWatch wrote that the terms were not disclosed and that TASER was building its Axon AI division on Dextro’s technology. He said: “We’ve known TASER CEO Rick Smith and the Axon team for a long time,” and “The acquisition is a natural fit for both companies.” His advice to New York founders was: “Focus relentlessly on building products your end customers want.”⁠[5]

Introducing him at the start of a Latent Space episode, the host swyx said that he had started a company in college and that this was his route to being acquired into Axon, where he was a “director of AI.”⁠[3]

The Masters of Scale biography states that from 2017 to 2020 he was VP Engineering at OpenAI, overseeing teams in research, language, supercomputing, reinforcement learning, safety and policy, and helping ship GPT, CLIP and DALL-E; the 20VC episode page likewise says that at OpenAI he oversaw research on language, supercomputing, reinforcement learning, safety and policy, and that his teams shipped GPT, CLIP and DALL-E.⁠[9][10]

On Latent Space, asked whether he had been OpenAI’s 30th hire, he answered: “Yeah, 30, 35, something around there.” He said he joined OpenAI fairly early and led engineering there for about two and a half to three years; he recalled that Greg BrockmanGreg BrockmanGreg Brockman is a software engineer and entrepreneur, co-founder and president of OpenAI, and a former chief technology officer of Stripe. His work spans payment infrastructure, engineering organizations and large AI systems, alongside public discussion of AI development and governance.Open the full entry and Ilya SutskeverIlya SutskeverIlya Sutskever is a computer scientist who studied under Geoffrey Hinton, helped develop AlexNet and sequence-to-sequence learning, and is a co-founder and former chief scientist of OpenAI. As a board member in 2023 he took part in removing Sam Altman, and after leaving OpenAI in 2024 he founded Safe Superintelligence, which he has led as CEO since 2025.Open the full entry pulled him into a room to have him take over their direct reports, coalescing teams out of a couple of early initiatives.⁠[3][3]

In the same episode he recalled that OpenAI’s Dota effort was going hard at the time; that he spent a good deal of time sprinting to help get the GPT-2 paper out; that every day while GPT-3 was being scaled up he would wake up stressed; and that the supercomputing team run by Chris Berner was one of his teams. His post introducing Adept states that his team at OpenAI scaled the Transformer into GPT-2 and GPT-3, and Tech Brew writes that at OpenAI he helped ship models including GPT-2, GPT-3 and DALL-E.⁠[3][3][3][3][11][12]

On Latent Space he said that after OpenAI he led GoogleGoogleGoogle is an American technology company founded by Larry Page and Sergey Brin and incorporated in 1998. It became an Alphabet subsidiary in 2015. Its businesses include internet search, advertising, software platforms, cloud computing, and artificial intelligence.Open the full entry’s large-language-model efforts while also co-leading Google BrainGoogle BrainGoogle Brain was Google’s AI research team, co-founded by Jeff Dean and others in 2011. It open-sourced the TensorFlow machine learning system in 2015 and was merged with DeepMind into Google DeepMind in April 2023.Open the full entry as one of its leads, speaking of “my year where I led the Google LM effort.” The post introducing Adept states that he led Google’s giant model efforts; 20VC describes him as a co-lead of Google Brain; AI Business called him a Google research director in 2022, and GeekWire in 2025 likewise described him as a former Google Research director.⁠[3][3][11][10][13][14]

Tech Brew writes that most of Adept’s founders left Google around the end of 2021.⁠[12]

On Latent Space he said “we started the company in January 2022.” Tech Brew writes that Adept had been in operation since that January and that its founding team included seven Google Brain alumni; he was CEO, Ashish Vaswani chief scientist and Niki Parmar CTO. Adept’s introductory post states that its co-founders Ashish Vaswani and Niki Parmar invented the Transformer in 2017.⁠[3][12][11]

AI Business writes that Adept had nine founding members: Luan as CEO; Kelsey Schroeder, who had led product for Google machine learning; Erich Elsen, a GPU computing pioneer who co-led DeepMindGoogle DeepMindGoogle DeepMind is Google’s artificial intelligence research organization, formed in April 2023 by merging DeepMind, which Google acquired in 2014, with the Brain team from Google Research, and led by CEO Demis Hassabis. It developed AlphaGo, AlphaFold, and the Gemini family of models, and Hassabis and John Jumper received the 2024 Nobel Prize in Chemistry for AlphaFold’s protein structure prediction work.Open the full entry’s large-model scaling effort; the Google Brain research scientists Niki Parmar and Ashish Vaswani as CTO and chief scientist; Anmol Gulati, who built Google’s production speech-recognition model; the former Google software engineer Fred Bertsch; and Augustus Odena and Max Nye, who built Google’s code-generation model.⁠[13][11]

On April 26, 2022, he published “Introducing Adept” on Adept’s website, describing Adept as a machine-learning research and product lab “building general intelligence by enabling people and computers to work together creatively.” The post states that Adept was training a neural network to use “every software tool and API in the world” as a universal collaborator for knowledge workers, working alongside people inside existing software such as Airtable, Photoshop and Tableau, and argues that this interface-based approach is the most practical and safest path to general intelligence.⁠[11]

The post announced that Adept had raised $65 million from Saam Motamedi and Reid HoffmanReid HoffmanReid Hoffman is an American entrepreneur, venture investor, author and podcast host, co-founder of LinkedIn, Inflection AI and Manas AI, and a partner at Greylock. A former PayPal executive, he has built and backed businesses in professional networking, artificial intelligence and drug discovery. His co-authored books include The Startup of You and Blitzscaling.Open the full entry at Greylock, Lee Fixel at Addition, Root Ventures, and angel investors including Scott BelskyScott BelskyScott Belsky is an American entrepreneur, author and early-stage investor who co-founded the creative portfolio platform Behance; after Adobe acquired it in 2012 he served as Adobe’s chief product officer and chief strategy officer. Since 2025 he has been a partner at the independent film studio A24 and founder of A24 Labs, and since September 2026 also a Product Advisory Partner at Union Square Ventures (USV).Open the full entry, Howie Liu, Dara Khosrowshahi, Chris Re, Lip-Bu TanLip-Bu TanLip-Bu Tan is a Malaysia-born entrepreneur and venture capitalist who founded Walden International in 1987, served as chief executive officer of Cadence Design Systems from 2009 to 2021, and has been chief executive officer of Intel since March 2025.Open the full entry, Andrej KarpathyAndrej KarpathyAndrej Karpathy is an AI researcher and educator, one of the founding members of OpenAI, who served as Tesla’s Director of AI and led the Autopilot computer vision team from 2017 to 2022, founded the AI education company Eureka Labs in 2024, and joined Anthropic in 2026 to work on pre-training research; at Stanford he designed and taught the university’s first deep learning course, CS231n, and he coined the term “vibe coding.”Open the full entry, Nathan Benaich, David Ha, Jaan Tallinn, Sarah Meyohas and Kevin Kwok; Tech Brew calls the round a Series A.⁠[11][12]

Tech Brew writes that Adept set out to create an AI assistant for the layperson that would carry out tasks through hundreds of software tools such as Photoshop, Airtable and Twilio, and that it hoped to release initial demos and sign its first customers within a year; the article paraphrases his view that Adept’s product would differ from robotic process automation (RPA).⁠[12]

Adept’s blog lists “ACT-1: Transformer for Actions,” published on September 14, 2022. On Latent Space he recalled releasing the ACT-1 demo and said that Adept later pushed hard on work such as the Fuyu architecture; the episode page lists models Adept released, including Persimmon-8B, Fuyu-8B and Fuyu Heavy.⁠[15][3][3][3]

On March 14, 2023, Adept announced that it had raised $350 million in new funding in its Series B, led by General CatalystGeneral CatalystGeneral Catalyst (GC) is an investment firm founded in 2000 and headquartered in San Francisco, United States, created by Joel Cutler, David Fialkow and Bill Fitzgerald, that backs companies from the seed stage through growth.Open the full entry and co-led by Spark CapitalSpark CapitalSpark Capital is a U.S. venture capital firm founded in Boston in 2005, with offices in San Francisco, New York and Boston. Its early-stage and growth funds back technology businesses across fintech, enterprise software, AI and consumer products. Investments have included Twitter, Tumblr, Oculus, Plaid and Anthropic.Open the full entry, with participation from existing investors. CNBCCNBCCNBC is an American business and financial news cable channel launched by NBC in 1989; it merged with its rival FNN in 1991 and since January 2026 has been part of Versant Media Group, spun off from Comcast.Open the full entry wrote in 2024 that Adept was valued at more than $1 billion in early 2023 and was backed by Microsoft and NVIDIANVIDIANVIDIA is an American semiconductor and computing technology company that provides graphics processing units, artificial intelligence infrastructure, networking interconnects, and computing software, and is listed on Nasdaq under the ticker NVDA.Open the full entry.⁠[16][17]

The 20VC episode page of June 2024 states that he had raised more than $400 million for Adept from investors including Greylock, Andrej Karpathy, Scott Belsky, NVIDIA, ServiceNowServiceNowServiceNow is an American enterprise cloud software company founded by Fred Luddy in 2004. It initially entered the enterprise market through IT service management and later expanded its workflow platform to employee service, customer service, security, and artificial intelligence applications. The company listed on the New York Stock Exchange in 2012 under the ticker symbol NOW.Open the full entry and Workday.⁠[10]

On June 23, 2024, 20VC published an interview with him, introducing him as co-founder and CEO of Adept, a company building AI agents for knowledge workers. The episode title lists its topics: why foundation-model performance is not diminishing but the models are commoditising, why NVIDIA will enter the model space and models will enter the chip space, and the right business model for AI software.⁠[10]

On June 28, 2024, CNBC wrote that Amazon had hired him and other members of Adept’s team for its AGI team; he would lead an “AGI Autonomy” division and report to Rohit PrasadRohit PrasadRohit Prasad is an artificial intelligence executive who led work on Alexa and artificial general intelligence at Amazon and oversaw development of the Amazon Nova foundation model family; he has been chief executive officer of the robotics company Boston Dynamics since October 2026.Open the full entry, senior vice president and head scientist of Amazon’s AGI unit. In an internal memo whose contents Amazon confirmed, Rohit Prasad wrote that the company had hired him and “a few other deeply talented team members to our AGI team,” and that the deal “will accelerate our roadmap for building digital agents that can automate software workflows.” Amazon licensed Adept’s technology, multimodal models and some datasets on a non-exclusive basis.⁠[17]

Adept wrote the same day that its co-founders and some of its team were joining Amazon’s AGI organization and that Amazon was licensing Adept’s agent technology, multimodal models and a few datasets; Zach Brock, the head of engineering, took over as CEO, and Tim Weingarten remained head of product. CNBC wrote that Adept said building its own models would have required more capital, that the arrangement with Amazon let it focus on agents, and that it would continue as a standalone company.⁠[18][17]

Dow Jones Newswires wrote in 2026 that in the deal Amazon agreed to hire some of Adept’s employees and to pay about $330 million to license its technology, a figure that comes from The Wall Street JournalThe Wall Street JournalThe Wall Street Journal is a U.S. news publication founded in 1889, with a focus on business, finance and economics. It is published by Dow Jones, part of News Corp.Open the full entry. GeekWire wrote in 2026 that the U.S. Federal Trade Commission was said to have examined the deal in 2024, and that acqui-hires of this kind have drawn regulatory scrutiny amid concerns that tech giants use such structures to absorb AI start-ups without triggering formal merger reviews.⁠[19][20]

GeekWire writes that the San Francisco-based Amazon AGI Lab that he led was formally established in December 2024 as a small, focused research group, and that he was also Amazon’s vice president of Autonomy. In 2025 GeekWire gave his title as “vice president of AGI Autonomy.”⁠[20][14]

In March 2025 Amazon unveiled Nova Act, an AI model and developer toolkit for building agents that carry out tasks autonomously in web browsers, released as a research preview and the first release of the San Francisco AGI Lab. GeekWire writes that Nova Act came as a Python software package that takes natural-language instructions and lets agents run in the background, many at once and without displaying anything on screen, completing step-by-step browser tasks such as submitting time-off requests or placing recurring orders; he led the lab together with Pieter AbbeelPieter AbbeelPieter Abbeel is an artificial intelligence and robotics researcher from Belgium, a professor at UC Berkeley and director of its Robot Learning Lab, known for research on robot apprenticeship learning and deep reinforcement learning. He co-founded Gradescope and Covariant and joined Amazon with part of the Covariant team in 2024.Open the full entry, an Amazon Scholar in robotics.⁠[14][20]

At the launch he said: “We really think agents are the last missing piece on the path to general intelligence,” adding, “This is really meant to be a way to engage developers,” and “We want to see what developers are going to do with it.”⁠[14]

GeekWire writes that Nova Act became generally available as an Amazon Web ServicesAmazon Web ServicesAmazon Web Services (AWS) is the cloud computing business of Amazon. Launched in July 2002, it made its public debut with the Amazon S3 storage service on March 14, 2006, and has since added hundreds of services, including Amazon EC2; its segment net sales were about $46.0 billion in the third quarter of 2026.Open the full entry (AWS) service at re:Invent in December 2025. He said Nova Act had been adopted by customers including Hertz, 1Password1Password1Password is an identity security software company headquartered in Toronto, Canada (contracting entity AgileBits Inc.). It was founded in 2005 by Dave Teare and Roustem Karimov, built its business on the password manager of the same name, and in October 2026 announced that its annual recurring revenue had surpassed $500 million.Open the full entry and Amazon.com and had reached the top of the REALBench agent research leaderboard. Late that year Peter DeSantis, a senior vice president with 27 years at Amazon, took over the broader organization that includes the AGI team.⁠[20]

On Tuesday, February 24, 2026, he announced in a LinkedInLinkedInLinkedIn is a social networking platform for professionals. It was founded in 2002 by Reid Hoffman and others, launched in 2003, and listed on the New York Stock Exchange in 2011; after Microsoft acquired it for approximately $26.2 billion in 2016, it became a wholly owned Microsoft subsidiary. Its revenue comes from membership subscriptions, advertising, and recruitment solutions.Open the full entry post that he would leave Amazon at the end of the week to “cook up something new,” less than two years after joining. He wrote: “There’s incredible work to be done at Amazon and opportunities for me to take on more areas,” and “But with AGI so close, I decided to spend 100% of my time on teaching AI systems brand new capabilities”; GeekWire writes that he also said: “I have a bet for what’s next.”⁠[19][20]

GeekWire writes that in his post he said the team “will be in great hands” with Peter DeSantis, and that Amazon declined to comment beyond his post. Based on a review of LinkedIn profiles, GeekWire counted that four of the five Adept co-founders who had joined Amazon in the deal had by then left.⁠[20]

Digital Trends wrote in May 2011 that he had begun his studies at Andover and Yale University at the age when most people start middle school.⁠[2]

Digital Trends also wrote that he had been an investment associate with the Yale College Student Investment Group and a research intern at iRobot, the company that makes the Roomba robotic vacuum.⁠[2]

The Masters of Scale biography states that he holds a BS in Applied Math and Political Science from Yale University.⁠[9]

Latent Space published an interview with him titled “Why Google failed to make GPT-3 + why Multimodal Agents are the path to AGI,” hosted by Alessio and swyx; the show notes link to a video of Dextro’s presentation at Data Driven NYC in 2015.⁠[3]

Masters of Scale’s “AI + You” episodes “5 ways to ethically build — and use — AI” and “5 steps for impactful experimentation” list him as a featured speaker alongside Mustafa SuleymanMustafa SuleymanMustafa Suleyman is a British AI entrepreneur and CEO of Microsoft AI who co-founded DeepMind and Inflection AI.Open the full entry and Fei-Fei LiFei-Fei LiFei-Fei Li is an American computer scientist, a professor of computer science at Stanford University and founding co-director of the Stanford Institute for Human-Centered Artificial Intelligence, and the co-founder and CEO of the spatial intelligence company World Labs. She led the creation of ImageNet and has served as a vice president at Google and chief scientist of Google Cloud, as well as an independent director of Twitter. Since 2026, she has served as senior advisor on artificial intelligence to Stanford’s president and co-chair of the HAI advisory council.Open the full entry, introducing him as CEO and co-founder of Adept.⁠[9]

In 2026 Primitive Labs, a company that uses AI to model customer behavior, launched with Rohit Talluri, who had worked under him in Amazon’s Autonomy Lab, as CEO; GeekWire writes that the backers listed in the company’s launch post included him as an angel investor.⁠[21]

Thiel Fellowship: Official website: https://thielfellowship.org/ (opens in a new window).⁠[22]

OpenAI: Official website: https://openai.com/ (opens in a new window).⁠[23]

Google: Official website: https://about.google/ (opens in a new window).⁠[24]

Adept: Official website: https://www.adept.ai/ (opens in a new window).⁠[15]

Amazon: Official website: https://www.aboutamazon.com/ (opens in a new window).⁠[25]

X: Public account: https://x.com/jluan (opens in a new window).⁠[26]