Nikita Mishra


About Me

Here is the link to my new website.

I finished my Ph.D. at the Department of Computer Science in the University of Chicago under Prof. John Lafferty and Prof. Hank Hoffmann! Earlier, I finished my Integrated Bachelors and Masters from the Department of Mathematics, IIT Kharagpur.

I am currently building new machine learning systems at Twitch as Applied Scientist in the Search and Discovery team.

My research interests include Machine Learning and Statistics. In the past few years I have developed specialization around NLP, Information Retrieval and Search Engines and deployed several machine learning systems to production. Additionally, I have also worked on Deep Learning, Bayesian Models, Differential Privacy and Multi-Task Learning.

Please find my resume here.

Email: nmishra at cs dot uchicago dot edu


My recent news!

  • I participated at panel discussion about differential privacy at BigID SF Engineering Privacy Meetup hosted by Google on February 6, 2019.
  • I gave a talk at hosted by Google, Sunnyvale on January 30th, 2019.
  • I joined Twitch as Applied Scientist on January 28th, 2019!
  • Our ASPLOS 2018 paper will be recognized as an Honorable Mention in the IEEE Top picks as its May/June 2019 issue.
  • Research Papers

  • Proteus: Language and Runtime Support for Self-Adaptive Software Development - S Barati, FA Bartha, S Biswas, ... Nikita Mishra (Alphabetical list of authors) in IEEE Software, 2019.
  • Controlling AI Engines in Dynamic Environments - Nikita Mishra, Connor Imes, John D. Lafferty, Henry Hoffmann, in (SysML-2018), February 2018.
  • CALOREE: Learning Control for Predictable Latency and Low Energy. - Nikita Mishra, Connor Imes, John D. Lafferty, Henry Hoffmann, International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS-2018).
  • Memory cocktail therapy: a general learning-based framework to optimize dynamic tradeoffs in NVMs. Deng, Zhaoxia, Lunkai Zhang, Nikita Mishra, Henry Hoffmann, and Frederic T. Chong. IEEE/ACM International Symposium on Microarchitecture (MICRO-2017).
  • ESP: A Machine Learning Approach to Predicting Application Interference. Nikita Mishra, John D Lafferty, Henry Hoffmann - IEEE International Conference on Autonomic Computing (ICAC-2017).
  • (Nearly) Optimal Differentially Private Stochastic Multi-Arm Bandits - Nikita Mishra, Abhradeep Thakurta, International Conference on Conference on Uncertainty in Artificial Intelligence (UAI-2015) [Supplementary] [Poster] [Slides]
  • A Probabilistic Graphical Model-based Approach for Minimizing Energy Under Performance Constraints - Nikita Mishra, Huazhe Zhang, John D. Lafferty, Henry Hoffmann, International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS-2015), [Poster] [Slides]
  • Private Stochastic Multi-arm Bandits: From Theory to Practice- Nikita Mishra, Abhradeep Thakurta, in ICML 2014 Workshop on Learning, Security and Privacy.
  • Unsupervised query segmentation using only query logs- Nikita Mishra, Rishiraj Saha Roy, Niloy Ganguly, Srivatsan Laxman, and Monojit Choudhury, Twentieth International World Wide Web Conference (WWW 2011) [Poster]

  • Patent

  • Apparatus and method for optimizing quantifiable behavior in configurable devices and systems Nikita Mishra, John D. Lafferty, Henry Hoffmann, in US Patent App. 15/457,743.

  • Doctorate Thesis

    Statistical Methods for Improving Dynamic Scheduling and Resource Usage in Computing Systems, Doctor of philosophy in Computer Science
    Advisor: Prof. John Lafferty and Prof. Henry Hoffmann, University of Chicago, 2017
    This thesis is about using statistical methods for performance and power estimation which would allow us to develop better scheduling algorithms and also more energy efficient systems. In many deployments, computer systems are underutilized–meaning that applications have performance requirements that demand less than full system capacity. Ideally, we would take advantage of this under-utilization by allocating system resources so that the performance requirements are met and energy is minimized. This optimization problem is complicated by the fact that the performance and power consumption of various system configurations are often application–or even input–dependent. Thus, practically, minimizing energy for a performance constraint requires fast, accurate estimations of application-dependent performance and power tradeoffs. We propose a set a algorithms for Improving Dynamic Scheduling and Resource Usage in Computing Systems

    Masters Thesis

  • A Statistical Approach for Minimizing Energy under Performance Constraints, Masters in Computer Science
    Advisor: Prof. John Lafferty and Prof. Henry Hoffmann, University of Chicago, 2015
    We explore the problem of minimizing the energy consumed by a new application subject to some power constraint; e.g., real-time or quality-of-service requirement. Specifically, we study the problem of online identification of power and performance tradeoffs. We use a hierarchical Bayesian network to model this problem which essentially strengthens the estimators of power/performance of other application by using the data from applications.
  • Clustering and Classification and Techniques for Directional Datasets, Integrated Masters in Statistics and Informatics [Best thesis award]
    Advisor: Prof. Somesh Kumar, IIT Kharagpur, 2012
    We develop clustering and classification techniques for directional datasets. The statistics involving directional datasets are significantly different from the usual linear datasets since an order-statistics cannot be defined in this case. We look into the clustering problem, we propose a generative model for clustering of the combination of linear and circular data.

  • Reports

  • Segments in Web Search Queries: What? How? Why? - Rishiraj Saha Roy, Nikita Mishra, Niloy Ganguly, Srivatsan Laxman, and Monojit Choudhury , 2010

  • Other Projects

  • Learning to Design Cost Sensitive Diagnostic Policies
    Supervisor: Prof. Alex Gray, Georgia Institute of Technology, 2011
    The process of diagnosis is a decision-making process in which the diagnostician performs a sequence of tests culminating in a diagnostic decision. The project describes the problem of learning diagnostic policies from training examples. The optimal policy must perform diagnostic tests until further measurements do not reduce the expected total cost of diagnosis. The main idea behind our method is that instead of finding all the possible combination for the tests and then pruning down the search space to find the best sequence of tests; we go the other-way round. We first find the good combination of tests for different cost constraints, using logistic classifier with weighted L1 regularization, then try to sequence the various linear classifiers that we obtain.
  • Taxonomy of Web-Queries
    Supervisors: Dr. Srivatsan Laxman and Dr. Monojit Choudhury, Microsoft Research, 2010
    The project was based on constructing a general grammar for web-queries that essentially emphasizes on the role of the words in a query. We defined query units as intent phrases that either guide the search engines for finer results or help reorder the retrieved pages; and content phrases which are used to define the content that must be present in the web-page.