About

We Attract Doers Who Solve Practical Problems

CMU has spent decades building a culture where people care about using technology to solve real problems. More than half a century ago, Allen Newell and Herb Simon had a vision for a general problem-solver for the human race. Since then, their vision has become a filter: people attracted to building solutions to real-world problems come here. The result? One of the world's largest collections of people determined to build things that help humans. That's why we're the right place to lead the AI charge in the 21st century. If you missed our first all-hands meeting in early November, you can watch it on our YouTube channel.

AI Stack

We define artificial intelligence pretty simply: AI must understand the human needs and it must make smart design decisions based on that understanding. Despite the simple definition, though, AI isn't just one thing. It's a giant thing, built from technology blocks we call the AI Stack. At Carnegie Mellon, we view it as a toolbox — each block houses a set of technologies that scientists and researchers can reach for as they work on new initiatives. Expertise in all areas? Not required. Instead, we believe you can focus on one area and draw on other parts of the stack for help. Each block depends on the other for support. And AI endeavors that ignore parts of the stack won't succeed. 

Click on each item in the stack for a quick overview and links to CMU faculty members working in that area. 

  • Autonomy

    AI technologies at this level focus on creating systems that make their own decisions without human intervention. These systems solve problems when humans can't. For example, robots can search through rubble for disaster survivors, or sensors in self-driving cars can respond more quickly to impending accidents than human drivers.

    • Henny Admoni

      My work looks at automous and shared automy robots for assistive interactions (e.g., tutors, physical assistants).

    • Lujo Bauer

      My work includes research on automous software vulnerability detection.

    • Timo Baumann

      I work on spoken dialogue systems infrastructure.

    • David Brumley

      Research in the Autonomy layer includes cyber-autonomy, automous cyber defense and automous cyber offense.

    • Jon Cagan

      I study automous decision-making.

    • Howie Choset

      I am working on an architecture that combines sensing, plannning and acting at mulitple layers — especially for systems traveling in challenging terrains. I also do a lot of path planning work and search.

    • Nicolas Christin

      My work in the Autonomy layer focuses on automatic predictions of actively exploited vulnerabilities.

    • Roger Dannenberg

      I work on autonmous agents for collaborative music performance.

    • David Garlan

      I have two decades of research invested in self-adaptive systems that take responsibility for managing uncertain environments, changing requirements, faults and attacks. This work has been applied in areas such as resilient enterprise systems, self-securing systems and most recently mobile robots.

    • Martial Hebert

      My research in this area includes work with ground robots and drones.

    • David Held

    • Jan Hoffmann

    • Ralph Hollis

      I work on indoor, dynamically stable mobile robots.

    • Aaron Johnson

      My work in this area includes automous navigation, planning and task execution for robotics.

    • George Kantor

      We are developing automous mobile robots for agriculture.

    • Kris Kitani

    • Christopher Langmead

      I work in automated science (integrated experiment selection and execution).

    • Anthony Lattanze

      I work with small bots, some swarming.

    • Lori Levin

      I work on computational linguistics and language techlogies, for all languages, not just English.

    • Maxim Likhachev

      I develop real-time algorithms for reasoning, planning and learning in automous systems ranging from humaid robots and ground and aerial vehicles to robotic manipulators operating in household and manufacturing domains.

    • Simon Lucey

      My team works with automous drones and cars.

    • Brandon Lucia

      By eliminating batteries and instead harvesting energy, our work on platforms for intermittent deep neural netwoks enables devices to operate automously.

    • Ruben Martins

      I develop verified code for automous drones.

    • Jean Oh

      My research includes efforts in automous vehicles (ground & aerial) and navigation.

    • Andy Pavlo

      I work on self-driving databases.

    • Andre Platzer

      I'm excited about expanding my work in this area.

    • Ariel Procaccia

    • Deva Ramanan

      I have a growing program in automous vehicles.

    • Mahadev Satyanarayanan

      I have pioneered the concept of edge computing — using cloudlets that provide low-latency, high-bandwidth compute capability that is resilient to internet failures. This allows continued operation of edge components even under conditions of Internet failure, cloud inaccessibility, cyber attacks on cloud services, etc.

    • Sebastian Scherer

      My work in the layer includes research on automous drones.

    • Jeff Schneider

      I work on self-driving cars, focusing on learning algorithms to execute onboard and to model and predict system performance offline.

    • Vyas Sekar

      We have a number of efforts in software-defined network security that tackle the issue of automous/self-healing in network security.

    • Aaron Steinfeld

      Our work on socially appropriate robot motion includes automy. My team also explores automous robot functionality to support people with disabilities in complex buildings and transit stations.

    • Katia Sycara

      I work on automous single and multi-robot systems, and robotic swarms of UAVs and UGVs.

    • David Touretzky

      I create software for programming intelligent automous mobile robots.

    • Manuela Veloso

      My research includes work on automous mobile service robots, automous robot soccer teams, automous interaction robots, automous robots for autism therapy, automous robots for amaly detection, automous robots for learning of vitals of the environment, symbiotic automy in which robots ask for help and learn from such help, and learning by instruction and correction.

    • David Wettergreen

      I work on automous planetary rovers for scientific investigation.

    • John Zimmerman

      Mostly, my work focuses on acceptance and explanations from automous systems.

  • Human AI Interaction

    When we create artificial intelligence in this part of the stack, we're augmenting what humans can do. These technologies make our lives easier. One good example of work in this area is robotic arms attached to motorized wheelchairs that people with spinal cord injuries can direct with their gaze. Another exciting example is research that allows a computer system to interview a patient remotely to determine if they're depressed and need human intervention.

    • Umut Acar

      I work on building an infrastructure for education and instruction. The goal of this system is not purely AI related, the information we collect about the student's interaction and use of data can offer an excellent foundation and data for using AI techniques in improving learning and, more broadly, education.

    • Henny Admoni

      My primary area of research is socially and physically assistive robotics.

    • Lujo Bauer

      I develop end-user interfaces for online decision-making.

    • Timo Baumann

      My work in this area includes spoken dialogue systems and perceived quality of speech based on prosodic features.

    • Jon Cagan

      My research includes studying interactions between humans and computers for problem solving and strategy design.

    • Iliano Cervesato

      Some of my work on "deductive spreadsheets" to support relational reasoning focused on usability.

    • Steven Chase

    • Howie Choset

      I believe human-robot interaction has three parts: 1) collect the data, 2) synthesize the data (fuse, filter, etc.) and 3) present the data. I have a sense that a lot of HRI centers on item 3. My group is interested in item 2.

    • Nicolas Christin

      Work in this area looks at password creation guided by neural-network based password meters.

    • Roger Dannenberg

      I work on human-computer music performance, which aims to integrate computers into live performance by human musicians.

    • Afsaneh Doryab

      We're building an intelligent systems for treatment, monitoring and control of mental illnesses (bipolar disorder, depression), and persuasive and smart mobile techlogy to promote health and wellbeing.

    • Jodi Forlizzi

      I work on data visualization, and helping the public understand how AI works. My research also focuses on UX design and AI.

    • Robert Frederking

      Language work has always been a central component to AI as I understand it.

    • David Garlan

      I work in the area of "explanation," which allows an AI system to tell users why it is recommending a course of action.

    • Hartmut Geyer

      My work includes research on human-adaptive control of prosthetic legs.

    • Chris Harrison

    • David Held

    • Ralph Hollis

      My work includes physical HRI integrated with speech dialogue.

    • Jason Hong

      I develop intelligent agents for smartphones and user interfaces to help people make better privacy decisions.

    • Eduard Hovy

      I construct models of human characteristics to form profiles and interaction styles, and automated identification of the same from dialogues. Example projects include automated stereotype/profile creation from all people listed in Wikipedia, identification and discovery of person/role types in online support groups and Wikipedia editors, etc.

    • Kris Kitani

    • Niki Kittur

      I develop tools to help people invate faster and make better decisions through machine learning, AI, visualization and crowdsourcing.

    • Ken Koedinger

      I work in this area, both by creating interactive machine learning — whereby human experts can transfer expertise to machines by showing as much as by telling — and in creating intelligent tutoring systems, through which computers can transfer expertise to massive numbers of students.

    • Chinmay Kulkarni

      We have a project that investigates how humans interact with AI/game-theory-based reward rules.

    • Philip LeDuc

    • Golan Levin

      My work includes Interactive art and design.

    • Lori Levin

      I work on computational linguistics and language techlogies, for all languages, not just English.

    • Simon Lucey

      We have done work in facial landmark tracking and on guiding users on how to move their mobile device when attempting to gain 3D reconstruction of an object.

    • Bruce McLaren

      I work on AI support for ethical decision-making.

    • Florian Metze

      I design user interfaces for meeting support.

    • Brad Myers

      My group has worked for many years on what we call Programming by Demonstration and Demonstrational Interfaces, where we teach the computer what to do by giving it examples. We currently are working as part of the InMind project on teaching Smartphones how to perform new tasks by both conversational user interfaces and programming-by-example.

    • Graham Neubig

      Research in this area focuses on interfaces to assist simultaneous interpreters.

    • Amy Ogan

      We develop systems that support learners using AI.

    • Jean Oh

      I work on conversational mobile robots, which include efforts in natural language understanding and generation.

    • Barnabas Poczos

      My work in this area is most in question answering.

    • Ariel Procaccia

    • Bhiksha Raj

      I work in speech and speaker recognition.

    • Deva Ramanan

      My research includes a component that focuses on better extraction of human knowledge for training AI systems.

    • Kelly Rivers

      I investigate how students interact with feedback generated by an automatic system.

    • Carolyn Rose

      My research team has developed the Bazaar agent architecture in which AI agents act as facilitators of group work, enacting practices human facilitators have successfully used.

    • Roni Rosenfeld

      Our team has deployed telephone-based, speech-based multilingual systems for disseminating information to and collecting information from hard-to-reach low-literate people in developing countries.

    • Sebastian Scherer

      I work on human/swarm interaction.

    • Nihar Shah

      My work includes research in learning from people.

    • Yaser Sheikh

    • Mel Siegel

      My research includes work on human-computer interface concepts and devices.

    • Daniel Siewiorek

      My research in this areas includes developing Virtual Coaches (e.g., rehabilitation exercises for stroke patients, muscle training for prosthetics) and Cognitive Assistants (e.g. for voices operating medical equipment such as AEDs).

    • John Stamper

      Student-system interaction is a key component of our research in intelligent tutoring systems.

    • Aaron Steinfeld

      We do a lot of work in this area, ranging from task assistance, human trust in robots/AI, human interaction with failing automy, negative human actions (bullying, cheating, etc.), and embodiment. We also examine specific interactions, like how pedestrians interact with automous vehicles.

    • Katia Sycara

      I work on various aspects of HRI, most recently in human interaction with multi-robot systems and swarms, as well as human trust in automous systems and ethical decision making of automous systems.

    • Anthony Tomasic

      We modeling user interface activity to predict which options users will select in a transit information system user interface. We have also published on automatic form-filling based on an emailed request.

    • David Touretzky

      My Calypso robot intelligence framework makes the robot's representation of the world transparent to humans who interact with it.

    • Yulia Tsvetkov

      My research includes work on negotiation assistants and dialogue systems that build rapport.

    • Bogdan Vasilescu

      I study end-user programming, making programming more accessible for n-programmers through natural language interfaces for code synthesis.

    • Manuela Veloso

      I work on active learning of robot capabilities by experimentation and model refinement; explainable AI and explainable deep reinforcement learning; and learning from instruction.

    • Min Xu

      We plan to develop algorithms for interactive analysis of complicated Cellular Electron CryoTomography (CECT) data.

    • John Zimmerman

      The bulk of my work occurs in this layer. It really spans all types of users and all types of AI/ML systems.

  • Planning & Acting

    Systems in this part of the stack rely on optimization, safety, the knowledge network and strategic reasoning to make the best possible decision available and learn from the information researchers give them. Though slightly less sophisticated than systems employing the blocks at the very top of the stack, planning and acting technologies still rely on advanced systems and algorithms to positively impact the world. One great example of technology that falls into this category is the national kidney exchange —a sophisticated algorithm that matches potential kidney donors with people who need transplants.

    • Umut Acar

    • Timo Baumann

    • Iliano Cervesato

    • Fei Fang

    • David Garlan

    • Seth Goldstein

    • Mor Harchol-Balter

    • David Held

    • Christopher Langmead

    • Lori Levin

    • Simon Lucey

    • Brandon Lucia

    • Ruben Martins

    • Graham Neubig

    • Jean Oh

    • Frank Pfenning

    • Andre Platzer

    • Barnabas Poczos

    • Ariel Procaccia

    • R Ravi

    • Pradeep Ravikumar

    • Nick Sahinidis

    • Tuomas Sandholm

    • Sebastian Scherer

    • Vyas Sekar

    • John Stamper

    • Katia Sycara

    • Ameet Talwalkar

    • Anthony Tomasic

  • Decision Support

    This layer includes technologies that help humans make decisions. Where should 500 Lyft drivers be deployed, based on information we know about events and demand? How should emergency services be distributed after a disaster? Exciting examples of research in this area include work to identify instances of human trafficking, help locate victims, and collect and synthesize enough information that trends and patterns can be discovered and used to combat the problem.

    • Leman Akoglu

    • Lujo Bauer

      I work on supporting end-user decision-making in online settings (e.g., password creation).

    • Travis Breaux

      My decision support research looks at predictions of perceived privacy risks in information disclosures, and of design decisions for composable security.

    • David Brumley

      My work in this area looks at deciding where to spend analysis time to find vulnerabilities.

    • Jon Cagan

      My research includes work in human-computer problem solving, design and diagnostics.

    • Kathleen Carley

      We work on DoD applications, and have a few research efforts in health care.

    • Howie Choset

      My medical robotics research provides visualizaiton to surgeons performing minimally invasive surgery.

    • Nicolas Christin

      My research investigates predicting cybersecurity threats.

    • Afsaneh Doryab

    • Fei Fang

      Based on the work in ML layer, we have developed methods to suggesting ranger patrol routes to prevent poaching.

    • Jodi Forlizzi

    • David Garlan

      My research investigates the use of AI planning technology and game theory to support human operators in managing automous systems and provides new techniques that allow a system to explain its recommendations and actions to a user.

    • David Held

    • Eduard Hovy

      My research in this area includes web harvesting and information-extraction from the web and social media to support various decision-support projects funded by Department of Homeland Security and PriceWaterhouseCoopers.

    • George Kantor

      One goal of our research is to provide better decision-support tools for growers making management decisions and for plant breeders to make selection decisions.

    • Kris Kitani

    • Niki Kittur

      My work involves making sense of overwhelming information through crowdsourcing, machine learning and visualization.

    • Ken Koedinger

      Our Intelligent Tutor Systems is a form of decision support to provide adaptive aids to learners to optimize their learning outcomes, efficiency and felt rewards.

    • Chinmay Kulkarni

      We are researching how AI techniques can be combined with learning science to improve workforce training in stem cell culturing.

    • Christopher Langmead

      My work in this area includes research in clinical decision support.

    • Anthony Lattanze

      My work includes developing embedded controllers that can learn.

    • Philip LeDuc

    • Lori Levin

      I work on computational linguistics and language techlogies, for all languages, not just English.

    • Jian Ma

    • Ruben Martins

      My work includes discovering bugs in programs, and enforcing properties in aerial vehicles that will prevent drones/airplanes from crashing.

    • Florian Metze

      My research in this area includes extracting information from audio meetings.

    • Amy Ogan

      We use models to support decision-making, e.g., when to offer support to a student learning a topic.

    • Jean Oh

      My work includes efforts in human-robot team problem-solving, personal assistant and disaster response.

    • Andre Platzer

      I work on safety-critical decisions in cyber-physical systems.

    • Barnabas Poczos

      My work in this area is most in question answering.

    • Ariel Procaccia

    • Bhiksha Raj

      I work on time series modelling and prediction.

    • Deva Ramanan

      I have worked on DoD-supported applications that require biometric recognition.

    • R Ravi

      I work on decision-support systems for online retail and advertising systems.

    • Carolyn Rose

      We are developing a collaborative decision-support tool for patients and their caregivers who are making decisions about heart therapies.

    • Roni Rosenfeld

      My work in this area focuses on visualizing bio-surveillance and epidemiological forecasts in support of decision making by DoD and public health agencies.

    • Nick Sahinidis

      We develop business decision-making algorithms and work on a variety of applications for the chemical process industries.

    • Russell Schwartz

      We are working on making more accurate predictions of disease progression, specifically for cancers, with the goal of bringing these methods to clinical decision-making and trial planning.

    • Vyas Sekar

      We've had several recent research results on optimizing internet video and voice applications.

    • Michael Shamos

    • Mel Siegel

    • John Stamper

      Our predictive models are often deployed in systems that help with decision-making. One example is teacher dashboards to help educators recognize which students might have (or be having) difficulty.

    • Aaron Steinfeld

      I ran the user evaluation studies for the DARPA RADAR project, which focused on task support during white collar tasks.

    • Katia Sycara

      I have done work on identification of promotion campaigns in social media, as well as identification of arguments and argumentation tactics. These efforts could be useful for decision support

    • David Wettergreen

      My research includes work on information-theoretic planning and execution particularly for guiding robotic explorers.

    • Min Xu

      We extract information from Cellular Electron CryoTomography (CECT) data for supporting clinical diagnosis of complex diseases, such as cancer.

    • John Zimmerman

      I'm working on a decision support system for implanting mechanical hearts.

  • Modeling

    AI systems at the top of the stack rely on computer modeling to understand information. Models use computers to construct and manipulate abstract representations of situations and natural phenomenon in the world. For example, new research has allowed scientist to analyze photos of people to track their facial features and recognize their emotional states.

    • Henny Admoni

      Our research works on modeling human intentions so robots can assist them.

    • Leman Akoglu

    • Ziv Bar-Joseph

      Our research in this layer investigates modeling molecular biology systems.

    • Lujo Bauer

      I model users' security and privacy preferences using machine learning, and model attacker behavior using machine learning.

    • Timo Baumann

    • Taylor Berg-Kirkpatrick

    • Travis Breaux

      My work in this area models policies and regulations for reasoning about legal decisions.

    • Jon Cagan

      I work in data induction, and medical and other imaging.

    • Kathleen Carley

      Our work in this area revolves around network and agent-based models.

    • Steven Chase

    • Howie Choset

      I model dynamics of systems.

    • Roger Dannenberg

      I work on automatic music composition and learning music structure.

    • Afsaneh Doryab

      My work studies detecting and predicting the state of mental health (e.g., depression, bipolar, stress, sleep quality), and using a patient's phone usage to predicting the severity of symptoms iafter chemotherapy.

    • Fei Fang

      I have worked on modeling various security challenges using game theoretic framework, including high-profile infrastructure security, environmental conservation challenges such as anti-poaching, and cybersecurity.

    • David Garlan

      My research is developing user models that can be used by AI planning techlogy to produce human-system symbiotic automy.

    • Hartmut Geyer

      I develop models of how the human system controls and learns locomotion.

    • Matt Gormley

      My work includes modeling of human language, often by incorporating linguistics intuitions via graphical models.

    • Mor Harchol-Balter

      The type of modeling I do is called performance modeling. This involves creating mathematical models of computer systems to help us understand how to speed up their performance. For example, I model computer systems to understand how to improve scheduling, resource allocation, load balancing, caching and core allocation.

    • David Held

    • Jan Hoffmann

      I work on probabilistic programming languages.

    • Jason Hong

      My work models many things: relationships between people, human activity, well-being, urban settings and smartphone privacy behaviors

    • Eduard Hovy

      I work on the (semi-)automated creation of semantic models of complex world situations using text mining and extraction, coupled with ontologies and domain models, augmented by human antation. Example projects: DARPA-funded projects in AIDA and World Modelers programs.

    • Kris Kitani

    • Niki Kittur

    • Ken Koedinger

      Our work in this area involves creating computational models of human voices, experts, and the learning transfer transition between.

    • Christopher Langmead

      I model biological systems and disease processes.

    • Tai Sing Lee

      My work in this area focuses on model learning and computation in the visual cortex of the brain.

    • Golan Levin

      I work on hand-pose analysis and gesture recognition.

    • Lori Levin

      I work on computational linguistics and language techlogies, for all languages, not just English.

    • Simon Lucey

      We work on systems for modeling/reconstructing 3D objects from video/image sequences. Objects can vary (car, chair, human face), as can the the device (mobile phone or drone).

    • Jian Ma

    • Ruben Martins

      I model programs for verification and security purposes.

    • Florian Metze

      I'm creating new models for representing audio.

    • Graham Neubig

      Work in this area focuses on modeling language in various aspects: multilingual modeling, modeling of syntax, semantics, dialogue.

    • Amy Ogan

      We model cognitive, social, and affective states using educational data.

    • Andreas Pfenning

      We are applying convolutional neural network approaches to study gene sequence variation in the context of Alzheimer's disease and evolution.

    • Andre Platzer

    • Barnabas Poczos

      I work on scientific modeling (physics, chemistry, biology).

    • Ariel Procaccia

    • Bhiksha Raj

      My research includes efforts in speech recognition, content-based audio analysis and speaker biometrics.

    • Deva Ramanan

      I develop systems for understanding facial affects and body pose from visual measurements.

    • Pradeep Ravikumar

    • Carolyn Rose

      In online work contexts like Wikipedia, GitHub and Ravelry, we have modeled role-based behavior profiles and how they interact within collaborative groups to predict both group and individual outcomes.

    • Roni Rosenfeld

      I work on forecasting epidemics and estimating the current disease burden from multiple heterogeneous signals.

    • Russell Schwartz

      Much of our work requires developing models of biological processes and fitting them to experimental data of various kinds, again with different threads in gemics and biochemistry.

    • Vyas Sekar

      We have a number of projects in data-driven networking.

    • Yaser Sheikh

      I work on social modeling (e.g., what does the spatial geometry of a social group tell us about the intent of the members of the group?)

    • John Stamper

      Our work focuses on building models of student states, both from a cognitive and non-cognitive areas.

    • Aaron Steinfeld

      Our team is looking at how to enable robots to move through crowds in socially appropriate ways. This involves modeling and predicting human motions and groups.

    • Katia Sycara

      I work on deep learning for collaboration and competition in multi-agent societies

    • Ryan Tibshirani

    • Yulia Tsvetkov

      My research includes work in data-driven natural language processing: machine translation, semi-supervised learning for low-resource NLP models, computational ethics and NLP for social good.

    • Bogdan Vasilescu

      I create models of software development processes and teams; models of sustainability of open-source ecosystems; and bimodal models of natural language and source code.

    • David Wettergreen

      I study physical process modeling, such as geologic composition or biologic distribution using n-parametric, generative models.

    • Min Xu

      We extract information from Cellular Electron CryoTomography (CECT) data for multiscale modelling of the structural organization of cells.

  • Machine Learning

    Machine learning focuses on creating programs that learn from experience. It advances computing through exposure to new scenarios, testing and adaptation, while using pattern- and trend-detection to help the computer make better decisions in similar, subsequent situations. A relevant example of work in machine learning is using speech recognition technologies to identify the age, sex and location of the hoax callers that plague the U.S. Coast Guard.

    • Umut Acar

      I have done work on statistical inference on machine learning.

    • Henny Admoni

      Some of my work involves learning human activity from human behavior signals (e.g., what object people will manipulate next based on their eye gaze?).

    • Leman Akoglu

    • Ziv Bar-Joseph

      My research in this area looks at data integration.

    • Lujo Bauer

      My work includes adversarial machine learning, detecting software vulnerabilities through machine learning, and using machine learning to help users make security and privacy decisions.

    • Timo Baumann

    • Taylor Berg-Kirkpatrick

    • Jon Cagan

      My research in this area includes work in data abstraction, explainable AI and image analysis.

    • Kathleen Carley

      We use machine learning techniques and apply them to new data. We generally don't create new machine learning techiques.

    • Steven Chase

    • Howie Choset

      I am not an ML researcher, but use ML tools in my research.

    • Lorrie Crar

      Mostly I am interested in privacy issues and ethical issues related to AI.

    • Roger Dannenberg

      I work on efficient leaning algorithms for deep networks.

    • Afsaneh Doryab

      I model human behavior from data streams.

    • Fei Fang

      I have worked on (1) predicting wildlife poaching threats, (2) detecting objects of interest (e.g., human, wildlife) from UAV videos and (3) detecting illegal mining sites from satellite imagery.

    • Matt Gormley

      I work on machine learning for natural language processing. Some NLP topics include grammar induction, dependency parsing, semantic parsing, knowledge base population, coreference resolution, topic modeling and computational semantics. Machine learning examples include approximate inference, unsupervised learning, decomposition techniques, nparametrics, global optimization, low-resource learning and approximation-aware learning.

    • Martial Hebert

      I work on minimizing supervision in learning for vision systems.

    • David Held

    • Jan Hoffmann

      My research in this area includes a focus on program synthesis and work on an automatic algorithm designer.

    • Eduard Hovy

      I investigate methods for distributional semantics and deep learning to capture the semantics of text.

    • George Kantor

      We use multiple ML methods for perception (e.g., extracting relevant features from raw images) and, more recently, for finding associations between plant getype and phetype data.

    • Carl Kingsford

    • Kris Kitani

    • Niki Kittur

      I study machine learning and AI models to augment analogy and decision making.

    • Ken Koedinger

      We combine and advance ML, AI and HCI techniques to create computational models of how humans develop complex expertise (for example, in math, science or languages).

    • Christopher Langmead

      My work in this layer falls into the supervised and unsupervised realm. I also work in active learning, Bayesian methods and parameter estimation.

    • Anthony Lattanze

    • Philip LeDuc

    • Tai Sing Lee

      I'm working on deep learning in artificial and real neural networks.

    • Lori Levin

      I work on computational linguistics and language techlogies, for all languages, not just English.

    • Simon Lucey

      We work on how 3D geometry can best be combined with recent advances in deep learning. We also work in gaining alternate theoretic motivations for why deep learning has been so successful, with the hope of applying these new theoretical insights to new problems (e.g. unstructured point clouds, geometry, efficient deep learning on mobile devices).

    • Jian Ma

    • Florian Metze

      My work in this area includes sequence modeling for speech recognition.

    • Graham Neubig

      My work in the Machine Learning layer includes dynamic neural networks, deep learning, DyNet neural network toolkit, language modeling, active learning and unsupervised/semi-supervised learning.

    • Amy Ogan

      We run/optimize machine learning algorithms to detect patterns in educational data.

    • Jean Oh

      My work includes efforts in natural language understanding, image-to-text translation and visual question answering.

    • Andy Pavlo

      We have an auto-tuning service for database systems called OtterTune.

    • Andreas Pfenning

      We are adapting attention models, which have been developed for imaging analysis, to understand how different gene sequences influence disease predisposition.

    • Andre Platzer

      My work in this area focuses on provable safety of reinforcement learning.

    • Barnabas Poczos

      I work on machine learning on complex objects, optimization and hyper-parameter tuning.

    • Ariel Procaccia

    • Bhiksha Raj

      I work in sparse optimization, privacy and security.

    • Deva Ramanan

      I focus on perceptual understanding.

    • R Ravi

      My work in this area focuses on recommendation system diversity.

    • Pradeep Ravikumar

    • Kelly Rivers

      My research includes work on data-driven generation of next-step hints for programming problems, based on other student submissions on the same problem.

    • Carolyn Rose

      My research group has developed deep learning approaches to detection of rhetorical relations in scientific discourse that has doubled the state of the art.

    • Roni Rosenfeld

      My work includes modeling of jointly evolving time series.

    • Nick Sahinidis

      We develop new ML algorithms of the best subset selection type.

    • Sebastian Scherer

      My work in this area focuses on LiDAR and image data classification to enable safer automy.

    • Jeff Schneider

      I work in algorithms for Bayesian optimization, including methods for optimizing in high dimensions, utilizing multifidelity feedback, and batch execution on highly parallel compute clusters. I also work on algorithms for learning that have complex inputs and outputs, such as distributions, functions, graphs and mixtures thereof. Some of the algorithms are for deep learning while others use kernel-based approaches.

    • Russell Schwartz

      We are working on two areas of machine learning in biology. One is learning the structure of gemic data sets and using that for prediction of disease progression. Another is learning models of biochemical processes from various kinds of experimental data.

    • Nihar Shah

      My work includes research in learning from people.

    • Yaser Sheikh

      I work in applied machine learning for computer vision and speech processing.

    • John Stamper

      My research builds predictive models of student knowledge. With colleagues from CMU and elsewhere, we have built LearnSphere, a platform to share analytics for educational data.

    • Katia Sycara

      I work in deep learning with particular interests in deep reinforcement learning for robot/agent control, learning to collaborate and compete. I also work in transparency and explanation of deep learning networks.

    • Ameet Talwalkar

    • Ryan Tibshirani

    • Yulia Tsvetkov

      My research includes work in data-driven natural language processing: machine translation, semi-supervised learning for low-resource NLP models, computational ethics and NLP for social good.

    • Bogdan Vasilescu

      I work on statistical NLP models, such as machine translation, but applied to software source code. Example problems include source code deobfuscation (JavaScript, reverse engineering binaries).

    • Manuela Veloso

      I work on deep learning for object and scene classification, and new deep reinforcement learning for multiagent policy.

    • Larry Wasserman

    • David Wettergreen

      My work in this area includes machine learning applied to image analysis (including hyperspectral), geometric data modeling (including LiDAR) and various classification tasks.

    • Min Xu

      We work on developing image analysis and machine learning approaches for analyzing the complicated structural contents in Cellular Electron CryoTomography (CECT) 3D images.

    • John Zimmerman

      My work is between people and systems. At this level, the work often focuses on how to obtain labelled data.

  • Massive Data Management

    Machine learning focuses on creating programs that learn from experience. It advances computing through exposure to new scenarios, testing and adaptation, while using pattern- and trend-detection to help the computer make better decisions in similar, subsequent situations. A relevant example of work in machine learning is using speech recognition technologies to identify the age, sex and location of the hoax callers that plague the U.S. Coast Guard.

    • Umut Acar

      My work in this layer focuses on two areas: 1) Algorithmic and system support for efficient processing of large data sets via incremental computation. 2) Parallel algorithms and code for processing large graphs.

    • Ziv Bar-Joseph

    • Timo Baumann

      I work on the Spoken Wikipedia Corpus.

    • Kathleen Carley

      Our team focuses on social media analytics and network science.

    • Afsaneh Doryab

    • Greg Ganger

      My reserach in this area focuses on robust, efficient, large-scale data storage and management, and system support for efficient, scalable ML training and inference.

    • Eduard Hovy

      My research includes data analytics to discover patterns and trends (causal and other) over various kinds of massive data. Example projects: combating human trafficking, discovering causal relationships in time series.

    • Carl Kingsford

      My work in this layer involves research on large-scale genetic analysis.

    • Ken Koedinger

      We provide large-scale web-based data and analytics sharing resources for human learning data through LearnSphere.org.

    • Christopher Langmead

      My research involves work with Apache Spark and the Cassandra database.

    • Lori Levin

      I work on computational linguistics and language techlogies, for all languages, not just English.

    • Jian Ma

    • Florian Metze

      Research in this layer focuses on large-scale analysis of speech and video data.

    • Andy Pavlo

      We're building a new self-driving database management system called Peloton.

    • Barnabas Poczos

      My work in this area includes cosmological surveys, large simulations and generated handwritten text.

    • Bhiksha Raj

      My research includes work on fast retrieval from private data on the cloud (privacy).

    • Pradeep Ravikumar

    • Carolyn Rose

      My group has developed DiscourseDB, which is a central data repository to facilitate research related to conversational data at a massive scale. This is part of the larger LearnSphere project.

    • Majd Sakr

      I teach a few courses that deal with data management and processing.

    • Yaser Sheikh

      I work with large multiview systems.

    • John Stamper

      I manage DataShop, the largest open repository of transaction based educational data.

    • Ameet Talwalkar

    • Anthony Tomasic

      We apply machine learning analysis to incoming transactions in main-memory database systems to improve system performance.

    • Rashmi Vinayak

      I study resource-efficient fault tolerance and failure recovery in big-data systems.

    • Min Xu

      We work on 3D grey scale images of size 6000x6000x1500 voxels each.

    • Jean Yang

      I'm interested in private data analytics.

  • Devices

    The device layer is all of the sensors and components needed for machines to perceive the world around them. Traffic lights, for example can observe traffic levels and negotiate with each other to improve traffic flow. Iris-detection systems can read and identify a human eye from 40 feet away.

    • Nathan Beckmann

      I building infrastructures for neural network inference on energy-harvesting devices w colleagues in the College of Engineering. Our work fits into many layers of the stack, but is most relevant in the Device layer.

    • Howie Choset

      I work on both different mobility systems (e.g., snakes) and modular systems.

    • Afsaneh Doryab

      I study using smartphone and wearable sensors to detect and predict human behavior.

    • Gary Fedder

      I work on MEMS, wearable and implantable systems.

    • Hartmut Geyer

      I work on robotic prosthesis control to help amputees improve their mobility and quality of life.

    • Chris Harrison

    • Martial Hebert

      My work includes research on object recognition/scene understanding from cameras and LADARs.

    • David Held

      My work is in developing new computer vision algorithms and new vision sensors.

    • Ralph Hollis

      My work in this area includes fundamental sensing, actuation and control.

    • Aaron Johnson

      I engineer mechanical intelligence at the reflex or preflex level.

    • George Kantor

      I develop platforms and sensors to operate in agricultural environments.

    • Anthony Lattanze

    • Philip LeDuc

    • Lori Levin

      I work on computational linguistics and language techlogies, for all languages, not just English.

    • Brandon Lucia

      We are developing extremely low-power computer architectures, hardware systems and software systems for inference tasks. We build systems that operate intermittently, harvesting energy from their environment, eliminating the need for a battery or a tethered power supply. Our goal is to define the system stack — hardware and software — for deep neural network computations on intermittent energy-harvesting devices.

    • Srinivasa Narasimhan

      My work in this area includes building next-generation cameras that can see below the skin and discover objects in the line of sight. We're also building energy-efficient 3D video cameras for robots, cars and consumer electronics, for VR/AR and automous navigation.

    • Jean Oh

      I use physical robot platforms for my research.

    • Carolyn Rose

      My research focuses on detection of social processes and states as they are reflected through natural language.

    • Yaser Sheikh

      I design sensor systems to measure social signals.

    • Justine Sherry

      Networking is one of the critical bottlenecks in large-scale machine learning. Recent work on datacenter networks has shown that ML training workloads are the most impacted by poor network performance.

    • Mel Siegel

      I work on sensing, sensors, perception/3D display systems.

    • Peter Steenkiste

      My research is in networking (the internet, wireless).

    • Aaron Steinfeld

      Our team is conducting research on transit information systems using our Tiramisu transit app. This work includes machine learning functionality and user interaction logs.

    • Katia Sycara

      I have done extensive work in negotiation algorithms that could be applied at this layer.

    • Ameet Talwalkar

    • Anthony Tomasic

      We model user activity through mobile phone accelerometers to determine when transit buses are full.

  • Computing

    All artificial intelligence is built on the computer systems that came before it. This includes the systems, networks, programming languages, operating systems and interactions between devices that make computing possible.

  • Ethics

    Ethics permeates the entire AI stack. The decisions people make as they build AI systems involve serious ethical questions that we can't ignore. At CMU, we believe that a vital component of AI is giving tomorrow's scientists the tools they need to perform ethical reasoning and the skills to create AI for good.

AI Is in Our Blood

Self-driving cars. Speech recognition. Advanced sensor design and application. Internet of things. Facial recognition software. Understanding human goals, thinking and interaction. We do all of that. But we also recognize that AI rests on top of a huge stack that relies on machine learning, programming, data analysis, design, physics and math. Our excellence in all those areas is part of what sets CMU apart from AI programs around the world.

Explore Research Areas

Our Degrees Prepare You To Create the AI of the Future

We offer a host of degrees that will prepare you for creating the AI technology of the future. From our undergraduate degrees in CS and computational biology — which allow you to do some research — to our many master's and doctoral degrees. All of our programs give you the hands-on, problem-solving experience that sets a CMU degree apart from all others.

Learn More About Our Programs

Undergraduate Programs

Graduate Programs