Amnon Shashua
Last reviewed
Apr 28, 2026
Sources
40 citations
Review status
Source-backed
Revision
v1 ยท 3,211 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Last reviewed
Apr 28, 2026
Sources
40 citations
Review status
Source-backed
Revision
v1 ยท 3,211 words
Add missing citations, update stale details, or suggest a clearer explanation.
Amnon Shashua (born May 26, 1960) is an Israeli computer scientist, entrepreneur, and the Sachs Professor of Computer Science at the [[hebrew_university_of_jerusalem|Hebrew University of Jerusalem]]. He is best known as the co-founder, President, and CEO of [[mobileye|Mobileye]], the world's leading developer of vision-based [[adas|advanced driver-assistance systems]] (ADAS) and [[autonomous_vehicle|autonomous-driving]] technology, which [[intel|Intel]] acquired for approximately $15.3 billion in August 2017 and which returned to public markets through a second IPO on the Nasdaq in October 2022[1][2][3]. Shashua also co-founded [[orcam|OrCam Technologies]] (2010, assistive devices for the blind and visually impaired), [[ai21_labs|AI21 Labs]] (2017, generative AI), [[mentee_robotics|Mentee Robotics]] (2022, humanoid robots), the digital private bank ONE ZERO (2019), and AAI Technologies (2023). His academic work introduced the [[trifocal_tensor|trifocal tensor]] as a fundamental construct of [[multi_view_geometry|multi-view geometry]] and applied [[tensor_factorization|tensor factorization]] to learning[4][5].
In 2020 Shashua received the Dan David Prize in Artificial Intelligence, in 2023 the [[israel_prize|Israel Prize]] for Lifetime Achievement, and in 2025 he was named to the TIME100 AI list[6][7][8]. In January 2026 Mobileye announced the acquisition of Mentee Robotics for approximately $900 million, a move into general-purpose robotics that Shashua described as the start of "Mobileye 3.0"[9][10].
Shashua was born in Ramat Gan, [[israel|Israel]], to parents of Iraqi-Jewish origin. He attended the ORT vocational school in Givatayim where he studied computer engineering[2][4]. He earned a Bachelor of Science in mathematics and computer science from [[tel_aviv_university|Tel Aviv University]] in 1986, followed by a Master of Science in computer science from the [[weizmann_institute_of_science|Weizmann Institute of Science]] in 1989[4]. Between his undergraduate and graduate studies he worked at Israel Aerospace Industries.
In 1993 Shashua received a doctorate in Brain and Cognitive Sciences from the [[mit|Massachusetts Institute of Technology]], where he worked at the MIT Artificial Intelligence Laboratory under the supervision of Shimon Ullman[4][11]. His PhD thesis, Geometry and photometry in 3D visual recognition, addressed how rigid 3D objects can be recognised from a small number of two-dimensional images using projective and affine invariants. He then held a postdoctoral position at MIT's Center for Biological and Computational Learning under Tomaso Poggio[4][11].
Shashua joined the Hebrew University of Jerusalem as a faculty member in 1996. He was promoted to associate professor in 1999, full professor in 2003, and appointed to the Sachs Chair in Computer Science in 2007. From 2002 to 2005 he served as head of the Hebrew University's School of Engineering and Computer Science[4].
Shashua's academic work spans three decades and roughly 160 papers, with research interests in early visual processing, [[multi_view_geometry|multi-view geometry]], tensor methods, [[structure_from_motion|structure from motion]], visual recognition, and the mathematical foundations of [[large_language_model|deep learning]][1][12]. He has continued to supervise research at the Hebrew University throughout his entrepreneurial career.
His early papers on the trilinearity of three perspective views showed that the geometric relationship among three calibrated cameras viewing a static scene is captured by a 3-by-3-by-3 tensor, now known as the [[trifocal_tensor|trifocal tensor]][5][13]. The 1995 result, refined in Trilinear Tensor: The Fundamental Construct of Multiple-view Geometry and Its Applications, became a textbook concept in Hartley and Zisserman's Multiple View Geometry in Computer Vision[5][14]. Shashua followed this with work on the quadrifocal tensor for four views and self-calibration techniques used in 3D reconstruction.
A second strand of his research applied [[tensor_factorization|tensor factorization]] to machine learning. Papers with Tamir Hazan on non-negative tensor factorization (2005) and later collaborations with PhD students Nadav Cohen and Or Sharir produced the 2015 ICML paper On the Expressive Power of Deep Learning: A Tensor Analysis, which proved separation results between deep and shallow networks and helped formalise why depth contributes to representational capacity[12].
Shashua has supervised doctoral students who became prominent computer-vision researchers in their own right, including Lior Wolf and Tamir Hazan. He holds more than 45 patents in machine vision[1][4].
Shashua co-founded Mobileye with [[ziv_aviram|Ziv Aviram]] in Jerusalem on May 13, 1999, spinning the company out of his Hebrew University laboratory after demonstrations showed that a single windshield-mounted camera and bespoke software could detect vehicles, lane markings, and pedestrians reliably enough for driver-assistance use[1][3][15]. Aviram took the role of chief executive officer and Shashua served as chairman and chief technology officer.
The company's commercial breakthrough came with the EyeQ family of vision-processing systems-on-chip. EyeQ1, sampled in 2004 and reaching production around 2007 to 2008, used a 180-nanometre process and ran Mobileye's first generation of vehicle, lane, and pedestrian detection algorithms[16]. Successive generations followed: EyeQ2 (around 2010), EyeQ3 (2014), EyeQ4 (2018), EyeQ5 (2021), and the EyeQ6 family in 2023, with EyeQ6L ("Lite") targeting Level 1 and Level 2 driver assistance and EyeQ6H delivering roughly three times the compute of EyeQ5H[16][17]. By the mid-2020s Mobileye had shipped over 200 million [[eyeq|EyeQ]] chips to customers including [[bmw|BMW]], [[audi|Audi]], [[volvo|Volvo]], [[gm|General Motors]], [[honda|Honda]], [[nissan|Nissan]], [[ford|Ford]], Stellantis, and the [[volkswagen|Volkswagen]] Group[1][16].
Mobileye became one of the first suppliers of camera systems for [[tesla|Tesla]]'s original Autopilot, but the partnership unravelled in 2016 after a fatal crash in Florida involving a Tesla Model S using Autopilot[18]. Mobileye declined to extend its supply contract, with Shashua arguing publicly that Tesla had "pushed the envelope in terms of safety" by allowing hands-free use of what Mobileye considered a Level 2 driver assist[18][19]. Tesla countered that Mobileye objected to its in-house chip development. The dispute influenced subsequent industry debate over advertised autonomy levels.
Mobileye held its initial public offering on the New York Stock Exchange on July 31, 2014 (shares opening for trading August 1), pricing at $25 per share and selling 35.6 million shares to raise approximately $890 million. The listing valued the company at roughly $5.3 billion of equity and was the largest US flotation by an Israeli company at the time[15][20].
On March 13, 2017, Intel announced an agreement to acquire Mobileye for $15.3 billion, paying $63.54 per share in cash; the transaction closed in August 2017 and remained the largest acquisition of an Israeli company in history for several years[1][3][15]. Shashua became a senior vice president of Intel while continuing as president and chief executive of Mobileye, which absorbed Intel's automated driving group and expanded into full-stack [[self_driving_car|self-driving]] systems, mapping, and robotaxi services.
In September 2022 Intel filed to take Mobileye public again, and on October 26, 2022 the unit listed on the Nasdaq Global Select Market under the ticker MBLY. Mobileye priced 41 million Class A shares at $21 (above its initial range), raising approximately $861 million for Intel and giving the unit an initial valuation of roughly $17 billion; shares closed up about 38 percent on the first day[21][22]. Intel retained a controlling Class B stake. The 2022 valuation was well below earlier reports that Mobileye might list at $50 billion, reflecting the compression in autonomous-driving valuations.
Mobileye's product line evolved beyond chips into platforms and services. SuperVision, an eyes-on hands-off Level 2+ system, used the multi-camera EyeQ5 architecture and shipped first on Geely's Zeekr 001 in China[16]. Mobileye Drive targets Level 4 eyes-off self-driving. The Road Experience Management (REM) programme uses crowdsourced data from millions of EyeQ-equipped vehicles to build high-definition maps. Chauffeur combines SuperVision with redundant sensors for personal autonomous vehicles, and Mobileye operates a robotaxi programme with partners including Volkswagen and Holon[1].
In 2017 Shashua and Mobileye colleague Shai Shalev-Shwartz published On a Formal Model of Safe and Scalable Self-driving Cars, which introduced [[rss_responsibility_sensitive_safety|Responsibility-Sensitive Safety]] (RSS), a mathematical framework that formalises a duty-of-care interpretation of road law into rules covering safe distances, right of way, and proper response to dangerous situations[23][24]. RSS has been adopted as a basis for IEEE P2846 and embedded in regulatory discussions in Europe, China, and the United States.
In 2010 Shashua and Aviram co-founded OrCam Technologies in Jerusalem, applying the single-camera computer vision approach used at Mobileye to the problem of assisting people with visual impairments[25][26]. Aviram took the chief-executive role while Shashua served as chairman.
OrCam's flagship device, the OrCam MyEye, is a small wearable camera and speaker that attaches magnetically to the temple of a pair of eyeglasses. The finger-sized device uses on-board AI to read printed and digital text, identify previously stored faces, recognise products and bank notes, and announce them through a tiny near-ear speaker. OrCam MyEye runs entirely offline and stores no images in the cloud, addressing privacy concerns and unreliable internet conditions that users often face[26][27]. The product launched commercially in 2015 and was named one of TIME's Best Inventions of 2019.
OrCam later launched OrCam Read, a handheld device for people with dyslexia or low vision; OrCam Read 3 (2024) which integrates generative AI to summarise and answer questions about scanned documents; and OrCam Hear, a solution for the hearing-impaired that uses lip-reading and speaker-direction cues to isolate a target voice in a noisy environment[27][28]. The company's devices are sold in more than 50 countries and have cooperative agreements with the National Federation of the Blind and the American Council of the Blind in the United States.
In November 2017 Shashua co-founded AI21 Labs in Tel Aviv with Stanford computer scientist [[yoav_shoham|Yoav Shoham]] and entrepreneur [[ori_goshen|Ori Goshen]][29][30]. Shoham, professor emeritus at Stanford and co-author of Multiagent Systems, became co-CEO with Goshen while Shashua served as chairman. The founding thesis was that the next generation of AI value would come from natural-language understanding rather than perception, and the company set out to build [[large_language_model|large language models]] for enterprise use.
AI21's first consumer product, Wordtune, launched on October 27, 2020 as a browser-based writing assistant that suggests rephrasings, expansions, and shortenings of user text. Wordtune attracted millions of users and was named one of Google's favourite Chrome extensions of 2021[29][30].
In August 2021 the company released [[jurassic_1|Jurassic-1]], a family of foundation language models that included a 178-billion-parameter version positioned as a direct competitor to OpenAI's [[gpt_3|GPT-3]]. Jurassic-1 used an unusually large 250,000-token vocabulary including multi-word expressions[29]. [[jurassic_2|Jurassic-2]] followed in March 2023 with improved instruction following, multilingual support, and lower latency, and AI21 Studio became the company's developer platform.
On March 28, 2024 AI21 announced [[jamba|Jamba]], a hybrid architecture combining [[mamba|Mamba]] state-space layers with Transformer attention layers and mixture-of-experts blocks; Jamba had 52 billion total parameters, 12 billion active per token, and a 256,000-token context window, making it one of the first production-grade SSM-Transformer hybrids[31][32]. Jamba 1.5, released in August 2024, scaled the architecture to a 398-billion-parameter Jamba 1.5 Large (94 billion active) and a 12-billion-parameter Jamba 1.5 Mini[32].
AI21's funding history matched its product cadence. After early seed and Series A rounds in 2019, the company raised a $64 million Series B in 2022, then closed a $155 million Series C round in August 2023 at a $1.4 billion valuation with backing from Google and Nvidia, and added an oversubscribed extension to bring the round to $208 million by November 2023[29][33][34]. A subsequent $300 million Series D was reported in 2024-2025, bringing total capital raised to roughly $636 million. Customers reported by AI21 include Atlassian, Salesforce, and Wix.
Shashua co-founded Mentee Robotics in 2022 to build general-purpose [[humanoid_robot|humanoid robots]] for domestic and industrial tasks[9][10]. The research approach emphasises simulation-first training, large-scale reinforcement learning in synthetic environments, and few-shot human-to-robot mentoring, with [[large_language_model|large language models]] used for high-level reasoning and task decomposition.
The MenteeBot prototype, demonstrated publicly in 2024, stands 1.76 metres tall, has 36 degrees of freedom, and uses bipedal locomotion with vision-based perception. Demonstrations showed the robot following spoken commands to set tables, fold laundry, and manipulate household objects. Mentee remained smaller than American competitors such as [[figure_ai|Figure AI]], [[1x_technologies|1X Technologies]], and [[apptronik|Apptronik]], but its tight integration of language, simulation, and control gave it a distinctive engineering profile.
In January 2026, during CES in Las Vegas, Mobileye announced an agreement to acquire Mentee Robotics for approximately $900 million (about $612 million in cash plus up to 26.2 million Mobileye Class A shares), with Shashua recusing himself from the Mobileye board's vote due to his founder stake[9][10]. Shashua described the acquisition as the start of "Mobileye 3.0". Mobileye plans to begin volume robot production with its automotive partner Aumovio in 2027, with household deployments targeted for 2030.
Shashua has founded or co-founded several additional companies. CogniTens, an industrial 3D-measurement firm based on stereo-vision techniques, was founded in 1995 and acquired by Hexagon AB in 2007[4]. ONE ZERO Digital Bank, an AI-first Israeli digital private bank, received a banking licence in 2019 and serves more than 100,000 customers; its conversational AI assistant Ella combines retrieval-augmented generation with bank-specific tools[35][36]. AAI Technologies, founded in 2023 with Hebrew University colleague Shai Shalev-Shwartz, develops what the founders call Expert AI, an architecture intended to combine deep neural reasoning with explicit symbolic knowledge graphs to address hallucinations in current [[large_language_model|large language models]][37].
Shashua's research output spans several areas of [[computer_vision|computer vision]] and machine learning that became standard references in the field.
| Area | Representative work | Year |
|---|---|---|
| Multi-view geometry | Trilinearity of three views, [[trifocal_tensor | trifocal tensor]] |
| Higher-order tensors | Quadrifocal tensor, projection matrices for higher-dimensional spaces | late 1990s |
| Visual recognition | Geometry and photometry in 3D recognition (PhD thesis) | 1993 |
| Tensor factorization | Non-negative tensor factorization (with Hazan) | 2005 |
| Tensor analysis of deep networks | On the Expressive Power of Deep Learning (with Cohen, Sharir) | 2015 |
| AV safety | [[rss_responsibility_sensitive_safety | Responsibility-Sensitive Safety]] (with Shalev-Shwartz) |
He is a frequent contributor to ICCV, CVPR, ECCV, NeurIPS, and ICML conferences. His dual identity as a working academic and the chief executive of a publicly traded company is unusual in the AI industry; he has stated that supervising graduate students keeps him connected to first-principles research.
Shashua has received major awards in academia, industry, and Israeli public life.
| Year | Award | Awarding body |
|---|---|---|
| 2026 | International member, US National Academy of Engineering | NAE |
| 2025 | TIME100 AI (Leaders) | TIME magazine |
| 2023 | [[israel_prize | Israel Prize]] for Lifetime Achievement |
| 2023 | Ramniceanu Prize for Engineering | Tel Aviv University |
| 2022 | Mobility Innovator Award | Automotive Hall of Fame |
| 2020 | Dan David Prize in Artificial Intelligence | Dan David Foundation |
| 2019 | Electronic Imaging Scientist of the Year | IS&T |
| 2019 | European Inventor Award finalist | European Patent Office |
| 2007 | Sachs Chair in Computer Science | Hebrew University of Jerusalem |
The 2023 Israel Prize cited his contributions to Israeli science, industry, and the global computer-vision community; the Dan David Prize, with a $1 million purse, recognised his role in establishing Israel as a leader in artificial intelligence[6][7].
Shashua is a regular speaker at CES, the World Economic Forum at Davos, and the IEEE Intelligent Vehicles Symposium. He has been outspoken about the need for transparent, model-based safety arguments in autonomous driving, contrasting the [[rss_responsibility_sensitive_safety|RSS]] approach with what he describes as opaque, learning-only systems[24]. In debates over [[tesla|Tesla]]'s [[elon_musk|Elon Musk]]-led Autopilot and Full Self-Driving programmes, Shashua has argued that consumer-facing self-driving features must be combined with redundant sensing, formal safety reasoning, and clear human-machine handover protocols, and he has cautioned against overheated timelines for general-purpose [[self_driving_car|self-driving cars]].
On AI more broadly, Shashua has expressed scepticism about claims that scaling [[large_language_model|large language models]] alone will produce general intelligence, arguing that hybrid architectures, formal reasoning, and curated knowledge will be needed; AAI Technologies and the [[jamba|Jamba]] line at AI21 are practical expressions of this view. Through the Shashua Family Foundation he and his family had donated approximately $60 million by 2024, including a $35 million COVID-19 relief fund (the WE-19 programme) and STEM scholarships at Tel Aviv University[1][4].