Dec 2019: Added many updates to my [ Research page ].
Nov 2019: Published Confident Learning: A family of theory and algorithms for characterizing, finding, and learning with label errors in datasets. Confident learning outperforms state-of-the-art (2019) approaches for learning with noisy labels by 30% increase an accuracy on CIFAR benchmarks with high label noise. [ paper | code | blog ]
Nov 2019: Announcing cleanlab: The official Python framework for machine learning and deep learning with noisy labels in datasets. [ code | docs ]
Jan 2019: Announcing the L7 blog! A place for machine learning and human learning. [ l7.curtisnorthcutt.com ]
See news for more.
My research focuses on two goals: (1) dataset uncertainty estimation, (2) the synergy of artificial intelligence to augment human intelligence. To this end, I established confident learning, a family of theory and algorithms for characterizing, finding, and learning with label errors in datasets, and cleanlab, the official Python framework for machine learning and deep learning with noisy labels in datasets. For an overview of my published research, please visit Google Scholar.
In addition to my MIT research, I am Chief AI Scientist at Knowledge AI, the principal author of the L7 machine learning blog, PomDP the PhD rapper, and a contingent research scientist at Oculus Research.
In my spare time, I help researchers build affordable state-of-the-art deep learning machines and enjoy competitive mountaineering, hiking, and cycling.
Some awards I’ve received include the MIT Morris Joseph Levin Masters Thesis Award, the NSF GRFP Fellowship, the Barry M. Goldwater National Scholarship, and the Vanderbilt Founder’s Medal (Valedictorian). I created the cheating detection system used by MITx and HarvardX online course teams, particularly in MicroMasters courses. At MIT, I TA’d 6.867, a 400-person advanced graduate machine learning course.
While these two ideas appear disparate, they are mutually dependent. Humans often have false notions about the world and encounter misinformation, yet we still learn well in noisy environments. Augmenting human learning with machine learning necessitates a deeper understanding of learning in noisy environments. Across healthcare, agriculture, politics, economics, transportation… our future as a species relies on an increasing synergy between machine learning and human learning: it is paramount that we have the tools to deal with real-world uncertainty, while maintaining the foresight to focus our advances in machine intelligence towards social good.
I am fortunate to have had the opportunity to work or intern at:
as well as academic collaborations and visiting research with MIT, Harvard, Vanderbilt, Notre Dame, and the University of Kentucky. Details here.
When you educate a person, you empower them within their community, and when you empower people socially, you give them hope, purpose, opportunity, and most importantly, you give them freedom.
Growing up below the poverty line in rural Kentucky, I experienced a glass ceiling of limited human and monetary resources. The ladder of opportunity often rises from prosperity rather than ability. My ladder was my education. Education led to exposure, then summer programs, then small scholarships, then bigger scholarships, and eventually opportunity. Everyone deserves access to quality educational resources – this underlies my motivation to pursue research that democratizes education.
To this end, I develop robust machine learning algorithms to enable open learning, i.e. to make advanced education more accessible. I work with edX student data to (1) infer user-intent across terabytes of noisy, massive interaction datasets and (2) implement prediction, inference, and detection algorithms distributed across 400+ MITx and HarvardX open online courses. For example, I ensure the legitimacy of online course certificates via cheating detection algorithms and with the help of exceptional colleagues, have demonstrated how machine learning can transform human learning with accurate proficiency estimation and diversification of comment rankings in discussion forums.
When asked if I like rap, I recommend PomDP the PhD rapper.