We work on both ML-in-the-large and ML-in-the-small. Our large scale ML work includes extreme classification (classifiers with millions of labels), and contextual recommendation systems (which recommend entities to users depending on context). Our work on contextual recommendation systems model situations in email clients or other productivity applications where there is no explicit query, but results are needed depending on implicit context of the user. Our research in ML-in-the small includes EdgeML, where we explore what it takes to run ML on small devices, and Natural Language Processing for low resource and mixed-code languages, where there is huge scarcity for labeled data. We also work on neuro symbolic reasoning, where we combine data driven machine learning methods with symbolic methods such as program synthesis, resulting in systems that are not only adaptive, but also interpretable and verifiable.
Some current projects
Understanding MixEd LANguaGE and Code-mixingThe goal of Project Mélange is to understand the uses of and build tools around code-mixing. Multilingual communities exhibit code-mixing, that is, mixing of two or more socially stable languages in a single conversation, sometimes even in a single utterance. This phenomenon has been widely studied by linguists and interaction scientists in the spoken language of such communities. However, with the prevalence of social media and other informal interactive…
Our objective is to develop a library of efficient machine learning algorithms that can run on severely resource-constrained edge and endpoint IoT devices ranging from the Arduino to the Raspberry Pi.
