AI for the robot age

Vicarious is developing artificial general intelligence for robots. By combining insights from generative probabilistic models and systems neuroscience, our architecture trains faster, adapts more readily, and generalizes more broadly than AI approaches commonly used today.

Research themes

The ability to generalize from a few training examples is one of the hallmarks of human intelligence. This ability is required for robots to work effectively in a variety of environments without arduous reprogramming.
Our algorithms learn models of the world that are then applied flexibly in a wide variety of situations. Our research emphasizes representations that enable task generality.
Underscoring our research strategy is the aim to discover the underlying properties of intelligence from neuroscience and cognitive science. We draw from the wealth of neuroscience literature to understand the representational structure and inductive biases that enable human brains to learn and generalize.
Teaching robots new tasks becomes easy if they have a conceptual understanding of the world just like humans. We develop algorithms that can learn abstract concepts from sensorimotor experience.


Prof. Fei-Fei Li

Dr. Li is the Director of the Stanford AI Lab, as well as an Associate Professor at Stanford. Prior to joining Stanford, she was on faculty at Princeton University and University of Illinois Urbana-Champaign. Research by Fei-Fei and her colleagues has been published in top-tier journals and conferences such as Nature, PNAS, Journal of Neuroscience, CVPR, ICCV, NIPS, ECCV, IJCV, IEEE-PAMI, and others. Fei-Fei is a recipient of the 2011 Alfred Sloan Faculty Award, 2012 Yahoo Labs FREP award, 2009 NSF CAREER award, the 2006 Microsoft Research New Faculty Fellowship and a number of Google Research aw

Prof. Bruno Olshausen

Professor Olshausen is the Director of the Redwood Center for Theoretical Neuroscience at UC Berkeley, as well as a Professor of the Helen Wills Neuroscience Institute. He serves on the Editorial Board of Vision Research and the Journal of Computational Neuroscience and was Chair of the Gordon Research Conference on Sensory Coding and the Natural Environment in 2004. In 2002, he co-edited the book Probabilistic Models of Perception and Brain Function (MIT Press).

Prof. Alan Yuille

Professor Yuille is the Director of the UCLA Center for Cognition, Vision, and Learning, as well as a Professor at the UCLA Department of Statistics, with courtesy appointments at the Departments of Psychology, Computer Science, and Psychiatry. He is affiliated with the UCLA Staglin Center for Cognitive Neuroscience, the Center for Brains, Minds and Machines, and the NSF Expedition in Computing, Visual Cortex On Silicon.

Our Team

D. Scott Phoenix

Before cofounding Vicarious, Scott was CEO of Frogmetrics (Y Combinator S2008), Entrepreneur in Residence at Founders Fund, CXO at OnlySecure (acquired by NetShops) and MarchingOrder (Ben Franklin Partners). Scott’s design work has been featured in 16 magazines and 3 museums, including the Institute for Contemporary Art in Philadelphia. Scott earned his BAS in Computer Science and Entrepreneurship from the University of Pennsylvania.

Dileep George, PhD

Before cofounding Vicarious, Dileep was CTO of Numenta, an AI company he cofounded with Jeff Hawkins and Donna Dubinsky. Before Numenta, Dileep was a Research Fellow at the Redwood Neuroscience Institute. Dileep has authored 22 patents and several influential papers on the mathematics of brain circuits. Dileep’s research on hierarchical models of the brain earned him a PhD in Electrical Engineerings from Stanford University. He earned in MS in EE from Stanford and his BS from IIT in Bombay.

Ken Kansky

Ken previously earned a BS degree in Computer Science and Physics at Stanford University, where he completed his thesis on image recognition under Peter Norvig. Before attending Stanford, Ken built laser control circuitry for condensed matter experiments at the University of Georgia, as well as a remote-control flamethrower, an electromagnetic levitator, and a flex sensor suit for controlling robotic arms.

Wolfgang Lehrach, PhD

Wolfgang was previously Director of Sequencing at Halcyon Molecular, where he developed image processing and bioinformatics algorithms to sequence DNA with electron microscopes. Before that, he worked at Alacris Theranostics, using deep sequencing and machine learning to enable personalized cancer treatment. Wolfgang did his post-doc in Machine Learning at Microsoft Research, earned his PhD and MSc from the University of Edinburgh, and his BS from Cambridge.

Bhaskara Marthi, PhD

Bhaskara Marthi was previously a Research Scientist at Willow Garage, where he devised algorithms for robots to build 3D maps of their surroundings, assemble Ikea furniture, and tidy rooms in an optimal way. Before that, Bhaskara completed his post-doc in artificial intelligence at MIT and earned his PhD in Computer Science from UC Berkeley. His favorite learning agent is two years old and has already achieved state of the art results on the task of detecting candy in cluttered images.

Charlotte Bowell, PhD

Before joining Vicarious as Director of Operations, Charlotte was a researcher at Integrated Plasmonics and Halcyon Molecular, where she manipulated and imaged the nanoworld. Before that we was a post-doc in the Electron Microscopy Group at the University of Cambridge. She first got a taste for manipulating sub-atomic particles during her PhD at the University of Birmingham in the Condensed Matter Group. Charlotte earned her MSci in Experimental and Theoretical Physics at the University of Cambridge.

Devin Gribbons

Devin previously earned his BA in Psychology at the University of Notre Dame, and his MFA in Fiction from the University of Alabama, where he was a Truman Capote fellow. He has worked in UX research and design at Verizon Laboratories and got his first experience coding at Harvard University’s Smithsonian Astrophysical Observatory. He was originally hired by Vicarious as an independent contractor and assisted with PR and office management before joining the team full time.

Huayan Wang, PhD

Huayan earned his PhD in Computer Science at Stanford University, under Daphne Koller. His thesis was on MAP inference methods in rich-structured graphical models. He has been a reviewer for NIPS, and author for ICML, AAAI, CVPR, and ECCV. He worked briefly at Coursera as a software engineer. Then he decided that building intelligent machines is what he really cares about. Before Stanford, he earned his M.E. and B.S. degree at Peking University.

Xinghua Lou, PhD

Xinghua Lou received his PhD. in Computer Science at Universität Heidelberg, under Fred Hamprecht. He worked on graphical models and structured learning for biomedical data analysis. He has published in NIPS, ICML, CVPR, Bioinformatics, IEEE-TMI, among others. Though initially trained as an experimental physicist in Engineering Physics at Tsinghua University, he is very excited about the current path towards building machines that can learn and reason.

Dennis Park, PhD

Dennis received his PhD in Computer Science from UC Irvine, where he was advised by Deva Ramanan. His research interests span computer vision and machine learning with a focus on visual recognition. His work on tracking people and their 2D body poses in videos was published in CVPR, ECCV, and ICCV. During his PhD, he worked as a research intern in Los Alamos National Laboratory and Microsoft Research, and was a recipient of ARCS award. He received his B.S in Physics from Seoul National University, South Korea.

Yi Liu, PhD

Yi was previously a Graduate Fellow at Stanford University, where his PhD was co-advised by Daphne Koller and Scott D. Boyd. His thesis work applied machine learning to the analysis of high-throughput sequencing of lymphocyte repertoires, and has been published in PNAS, Journal of Immunology, JAMIA, and others. Yi earned his BS in Math from University of Waterloo. In a past life, he worked in high-frequency trading and was also briefly one of the highest ranked Hearthstone players in North America.

Miguel Lazaro Gredilla, PhD

Miguel previously earned his PhD in machine learning at Universidad Carlos III de Madrid. In his thesis, Miguel developed a sparse Gaussian process model which has become the de facto benchmark for fast regression algorithms. Since then, Miguel has been working on approximate inference algorithms for other Bayesian models. He has been a visiting scholar at the University of Cambridge and the University of Manchester, and has authored and reviewed for NIPS, JMLR, and several IEEE journals.

Zhaoshi Meng, PhD

Zhaoshi (Josh) received his PhD from the University of Michigan, under the supervision of Alfred Hero and Long Nguyen. During his PhD, Zhaoshi has worked on probabilistic graphical models and co-authored the Best Paper at ICML 2014, a Notable Paper at AISTATS 2013, and a Best Student Paper at CAMSAP 2013. Zhaoshi earned his BS in EE from Tsinghua University, and worked on recommendation algorithms at Microsoft Research Asia. He enjoys discovering latent patterns from noisy observations.

Michael Stark, PhD

Michael was previously a member of the Max Planck Center for Visual Computing and Communication, both at Stanford University and MPII. He earned his PhD from TU Darmstadt for his research on knowledge transfer for object class recognition. Michael’s work has been published in CVPR, ICCV, ECCV, ICLR, and he received back-to-back Best Paper Awards at CVPR-3dRR. He is thrilled to be able to teach intelligent machines how to see.

Robert Hafner

Rob was previously the VP of IT at Malwarebytes, where he was responsible for all server side development and operations as well as the malware intelligence team. Prior to that he was the cofounder of SolunaNet, a Dev/Ops and Security consulting firm, where he worked with dozens of different clients to build scalable systems and security conscious developers. In his spare time he works on open source projects such as Stash (the caching library), JShrink (a javascript minifier), and Fetch (an IMAP client library).

Bryan Hart

Before joining Vicarious as its Resident Chef, Bryan worked as a private chef to Silicon Valley executives since 2001. He started his culinary adventure while living in Japan and working for the US Navy where he toured Asia and was exposed to new foods and cultures. After his return to the Bay Area, he attended the California Culinary Academy of San Francisco and graduated in 1994. Outside of work, Bryan spends his time with his children, as well as glassblowing and woodworking.

Nick Hay, PhD

Nick Hay earned his PhD at UC Berkeley under Stuart Russell. His research applied reinforcement learning and Bayesian analysis to the metalevel control problem: how can an agent learn to control its own computations. More broadly, he is interested in how AI systems can be safely developed for the benefit of humanity, and how this pursuit might be informed by the cognitive sciences. Born in New Zealand, Nick is still getting used to walking upside down.

Wenzhao Lian, PhD

Wenzhao received his PhD in machine learning at Duke University, advised by Dr. Lawrence Carin. He has worked on probabilistic graphical models and Bayesian statistics, focusing on time series and point processes. During his PhD, Wenzhao interned at Microsoft Research and Yahoo Labs; he has authored and reviewed for ICML, NIPS, AAAI, and IEEE journals. He earned his Master’s degree in statistics at Duke and Bachelor’s degree in ECE at Shanghai Jiao Tong University.

Sandhya Bhasker

Before joining the Operations Team at Vicarious, Sandhya worked in executive support and operations in both corporate and start-up environments. Channeling her previous experience in SEO, she expanded those roles to include marketing and brand-storytelling. Sandhya earned a BA in Neuroscience and History from Kenyon College, where she completed theses on traumatic brain injury and sub-Saharan mineral economies.

Roman Vasylenko

Roman was previously a senior system engineer at EPAM, where he was responsible for databases and applications deployment automations. Roman has also served as the lead consultant for a tech security team and worked as both a system and network administrator. Roman currently resides with his lovely wife and daughter and a big red cat named Oscar.

Anna Chen

Anna was previously at Quora, where she applied ML to learn the semantics of millions of questions, and built the first system to detect paraphrased questions. Before that she earned her MS in Computer Science at University of Michigan, where she worked with Edmund Durfee and Satinder Singh on multi-agent cooperative planning under a reinforcement learning framework. In her free time she likes to paint, currently focusing on watercolor and oil painting.

Scott Swingle

Scott earned his BS in Applied and Computational Mathematics at Brigham Young University. While there, he completed undergraduate research in handwriting recognition, created possibly the world’s best competitive bot for the game Puyo Puyo, programmed a cryptanalysis engine to attack a variety of common cyphers, and wrote a multi-threaded program to decipher a symmetric-key cryptography challenge. Scott loves to cook, solve online math challenges, and work on coding projects. He especially enjoys saving time by making computer programs that play online games autonomously, so that he doesn’t have to.

Tom Silver

Tom Silver earned his BA with highest honors in Computer Science and Mathematics at Harvard. His undergraduate thesis proposed a new crowdsourced benchmark for artificial intelligence. He previously interned at Numenta and Google. He also worked in the Sabeti Lab, publishing research on machine learning methods for Ebola prognostic prediction. Cooking, corgis, and coffee are a few of his favorite things.

Alex Schlegel, PhD

Alex completed his PhD in Cognitive Neuroscience with Peter Tse at Dartmouth College, where his work sought mechanistic links between the distributed networks of the brain and the mind they create. The questions that drive him are: How does the human brain mediate flexible, creative cognitive behaviors including artistic and scientific thought? Why is our species the only one (so far) to demonstrate these abilities? And how are these abilities realized via dynamic interactions in distributed neural networks? In his free time, Alex enjoys hiking, gardening, and creating sculpture that reveals hidden processes in the environment. He looks forward to the day when machines begin imagining a better world.

David A. Mely, PhD

David completed his Ph.D. in Cognitive Science with Dr. Thomas Serre at Brown University, elaborating computational models of cortical micro-circuitry to explain contextual integration in human vision. Previously, David’s interest in thinking machines was sparked by Isaac Asimov’s novels; he went on to study mathematics, theoretical physics, and biology at the École Polytechnique. When he is not working towards AI, David is a bon vivant who always enjoys craft beer, jazz music, and swimming.

Carter Wendelken, PhD

Carter was previously a research scientist at UC Berkeley where his work focused on neural mechanisms of reasoning, memory, and cognitive control. He has authored publications in journals including Cerebral Cortex, Neuron, Journal of Neuroscience, Developmental Science, and Neurocomputing. Prior to working as a neuroscientist, Carter earned his Ph.D. in Computer Science (Artificial Intelligence), under Lokendra Shastri and Jerry Feldman. Now that he has figured out how human brains work (if only), Carter is excited to apply what he’s learned to the task of creating artificial brains.

Mohamed Eldawy

Before joining Vicarious, Mohamed worked as a software engineer, first at Google and then at two startups. While at Google, Mohamed worked in the ranking team finding signals to improve web search. In his 20% time, he worked on content based image understanding. He earned a master’s degree from University of Wisconsin-Madison University of Wisconsin-Madison in 2007. Mohamed is deeply passionate about neuroscience, and spends his spare time learning about the inner workings of the brain. Outside of work, Mohamed enjoys running, swimming and learning new languages.

Nan Rong, PhD

Nan completed her Ph.D. in Computer Science with Dr. Joe Halpern at Cornell University, where her work focused on enabling machines to automatically learn new actions that they weren’t aware of previously. She also worked on designing equilibrium concepts that explain and motivate people’s cooperative behavior in a game theoretic setting. She is pretty serious about theories and at the same time excited by building robots. Her passion for creating robots started from reading Asimov’s Three Laws. In her free time, she enjoys reading novels, archery, and tree climbing.

Alex Lavin

Alex studied mechanical and aerospace engineering at Cornell, then pursued robotics in Carnegie Mellon’s MS in Mechanical Engineering program, where he led a team to building a lunar rover. Previously, he was a Senior Software Engineer with Numenta, developing biologically-derived AI algorithms. In 2016, Alex was selected to the Forbes 30 Under 30 List for Science. In his free time, Alex enjoys running, yoga, live music, and reading sci-fi and theoretical physics books.

Qinxun Bai, PhD

Qinxun received his PhD in Computer Science at Boston University, where he was advised by Stan Sclaroff and Steven Rosenberg, and also worked with Henry Lam. His PhD work was about the differential geometric structure in supervised learning of classifiers and Bayesian asymptotics for online ensemble learning. Qinxun earned his BS in optoelectronics from Tsinghua University and MS in pattern recognition from NLPR, both in Beijing. Outside of work, he enjoys outdoor sports and visual arts.

Ramki Gummadi, PhD

Before joining Vicarious, Ramki was a data scientist and software engineer at Facebook. Prior to that, he completed postdoc research at Stanford with Ramesh Johari in the Operations Research Group and at the University of Alberta with Dale Schuurmans in the RLAI group. He earned a PhD from the University of Illinois, with a dissertation on rateless erasure codes in communication networks, and a BS in Electrical Engineering from IIT Madras. In his free time, he enjoys running, reading, and the outdoors.

Jimmy Baraglia, PhD

Jimmy earned his Ph.D. in Engineering at the Osaka University under the supervision of Minoru Asada and Yukie Nagai. His work allowed robots to develop prosocial behavior similar to infants’ based on a prediction error minimization mechanism. The system is based on Bayesian inference and the Free Energy principle. Previously, Jimmy was student at Paul Sabatier University in France where he studied automatism, signal processing and electronics. Jimmy loves cooking, traveling and playing all kind of sports.

Eyrún Eyjólfsdóttir, PhD

Eyrún’s research interests lie on the intersection of Artificial Intelligence and Neuroscience. She completed her PhD at Caltech where she worked with Pietro Perona on automated analysis of behavior, particularly that of fruit flies, in a collaboration with the David Anderson lab. During her PhD, she interned at Qualcomm Research Center, where she developed a power-efficient context-classification algorithm for smartphones. Prior to Caltech, she received her MS in Computer Science from UCSB and her BS in Mathematics from the University of Iceland. Her favorite pastime activities involve the backcountry, home brewing, and knitting.

Jaldert Rombouts, PhD

Jaldert earned his PhD at the Free University of Amsterdam under Pieter Roelfsema and Sander Bohte for his work on biologically plausible learning in neural networks. At CMU’s Robotics Institute he worked on the SnackBot, and at Brain Corporation he helped develop indoor navigation solutions for large wheeled robots. Besides playing with state of the art learning systems, Jaldert likes cooking (especially breakfast, his favorite meal), biking, hiking, reading, and Krav Maga.

Nate Tucker

Nathaniel earned his AB/SM in Computer Science from Harvard. He previously worked as a Quant and Trader at Jane Street and Goldman Sachs before transitioning into the pure tech industry. Nathaniel worked as a Data Scientist at Facebook, a Product Manager at Microsoft and a Software Engineer at Google before joining Vicarious. He is an avid reader and learner. He teaches part time at General Assembly and is developing open source teaching material for data science, machine learning, and web development.

Amit Padhy

Before joining Vicarious, Amit worked at a healthcare software company building integrated systems for hospitals to help provide quality care for patients. Prior to that, he developed accurate and robust handwriting recognition algorithms with Thomas Binford at Read-Ink Technologies. Amit earned a B.Tech in Computer Science at IIT Madras where he became interested in intelligent systems while doing research in camera surveillance and object recognition. In his free time, Amit enjoys exploring new places and playing music.

Klas Kronander, PhD

Klas Kronander completed his PhD in Robotics at Ecole Polytechnique Federale de Lausanne under the supervision of Aude Billard. He has done research on robot learning from demonstration, specifically in the domains of compliant manipulation motions and dynamical systems. Outside of work, Klas enjoys spending time with his family, the outdoors, cooking and listening to jazz.

Dianhuan Lin, PhD

Dianhuan Lin received her PhD in Machine Learning at Imperial College London, advised by Stephen Muggleton. Her research focused on Symbolic Machine Learning, specifically Inductive Logic Programming. She developed approaches to automatically learn high-level abstraction, which leads to more compressed representation of concepts. She also worked with Josh Tenenbaum on one-shot program induction. Before joining Vicarious, she was at Amazon, working on Alexa. Outside of work, she enjoys all kinds of sports, including hiking, playing ping pong and swimming.

Matthieu Perrinel, PhD

Matthieu was previously a data scientist in Deepomatic, where he worked on the efficient building of similarity datasets, and designed deep learning based models to predict the similarity between objects. Before joining Deepomatic, Matthieu earned his Ph.D. in Computer Science at ENS Lyon, under Patrick Baillot. During his thesis, he created type systems for lambda-calculus corresponding to polynomial time. When he does not talk about french food, Matthieu enjoys sailing in the bay and brewing his own beers.

Christopher Tham

Before joining Vicarious as a Software Engineer, Chris earned his B.S. in Decision Science from Carnegie Mellon University. His technical interests lie in systems theory and data analytics, especially as applied to decision making in complex environments. In his free time, Chris enjoys playing poker and discovering good tunes.

Thomas Allen, PhD

Tom was previously a senior robotics engineer at Pneubotics, where he wrote control and estimation software for pneumatic robot arms.  Before that he earned his PhD in mechanical engineering from Caltech under Joel Burdick. As a part of a Caltech/JPL team he wrote grasp planning software for DARPA’s Autonomous Robotic Manipulation competition. In his free time, he climbs rocks.

Anjali Rao

Anjali Rao was previously a Robotics researcher at Symbio Robotics, where she worked on robotic learning for task generalization. Before that she earned her Master’s degree from SUNY Stony Brook University. Her Master’s thesis was focused on User-guided Inverse Kinematics based Path planning for redundant manipulators. She completed her Bachelors in Mechanical Engineering in India. Besides understanding the early stages of human brain development, she enjoys reading, traveling and cooking.

Levi Lovelock

Levi previously worked at Atlassian as a DevOps Engineer where he was responsible for infrastructure automation and continuous integration. He has a passion for “everything as code; cattle – not pets” when it comes to infrastructure. In his spare time, he enjoys making music on his guitar whilst also working on achieving his MS degree from Georgia Tech.

Swaroop Guntupalli, PhD

Swaroop completed his PhD in Cognitive Neuroscience at Dartmouth College under James Haxby. His work in the Haxby Lab focused on understanding neural representation of visual and auditory information and developing computational methods for probing and building models of representational spaces in the brain. Prior to Vicarious, he was a postdoc in HaxbyLab and a visiting scholar at Stanford. He received his MS in Computer Science from Texas A&M University under Bruce McCormick, and BS in Mechanical Engineering from IIT Madras. Swaroop enjoys running, hiking, reading math, science, and sci-fi, coffee, and hoppy IPAs.