Nimar Arora

Current Interests

Algorithms for Probabilistic Inference.
Probablistic Models of complex real-world processes.
Large-scale AI deployments.

Education

Ph.D., Computer Science, University of California, Berkeley, 2006 - 2012

Artificial Intelligence

M.S., Computer Science, University of Texas, Austin, 1996 - 1998

Distributed Systems

B.Tech., Computer Science, Indian Institute of Technology, Kanpur, 1992 - 1996

Work Experience

Head of AI and Analytics at DevRev, 2020 - 2023

AI, Analytics, and Automations Infrastructure

  • Built the AI, Analytics, and the Automations team from the ground up.
  • Platformized AI for easy extensibility.
  • Deployed many model such as Semantic Search.
  • Built the Analytics foundations to support all Go To Market teams.

Research Scientist at Facebook, 2018 - 2020

AI Infrastructure, Probability

  • Designing a new Probabilistic Programming Language (PPL).
  • Researching inference algorithms for PPLs.
  • Designing and deploying probabilistic models for various internal applications.
  • Designing an Open Source benchmarking tool for PPLs.

Founder at Bayesian Logic, 2012 - Present

  • Co-invented the NET-VISA generative model to help the CTBTO (a UN organization) detect and locate nuclear explosions as well as other seismic, hydro, and infrasound events on a global scale.
  • Deployed the NET-VISA model at the UN which includes real-time probabilistic inference as well as continuous retraining of the model.

Architect at Machine Zone (MZ), 2015 - 2017

Game Engineering, Data Science, Live Ops

Led a broad company-wide effort to automate in-app sales using Reinforcement Learning. This was deployed in "Game of War: Fire Age," "Mobile Strike," and "Final Fantasy XV: A New Empire."

Principal Member of Technical Staff at Oracle, 1998 - 2014

Distributed Databases, Replication

  • Architected Oracle Streams and later the integration with Oracle Golden Gate.
  • Invented and deployed various distributed and concurrent database replication algorithms.
  • Awarded 20+ patents.

Awards

Mitchell Prize

2014

Jointly awarded by the American Statistical Association and the International Society for Bayesian Analysis to recognize "an outstanding paper that describes how a Bayesian analysis has solved an important applied problem," for the paper "NET-VISA: Network Processing Vertically Integrated Seismic Analysis".

Outstanding Student Paper Award

2011

American Geophysical Union Fall Meeting, 2011 , for "Scalable Probabilistic Inference for Global Seismic Monitoring."

Skills

  • Machine Learning
  • Probabilistic Modeling
  • MCMC
  • Variational Inference
  • Deep Learning
  • Distributed Systems
  • Python
  • C++
  • Pytorch
  • Tensorflow/Keras
  • numpy
  • matplotlib

Conference Reviewer

  • NeurIPS (top-400 reviewer)
  • ICML
  • UAI
  • AAAI
  • AISTATS
  • ICLR

Selected Publications

  • Geeta Arora, Nimar Arora, et al.Importance Sampling-Based Estimate of Origin Error in NET-VISA, Pure And Applied Geophysics 2022.
  • Viet-An Nguyen, Peibei Shi, Jagdish Ramakrishnan, Narjes Torabi, Nimar Arora, Udi Weinsberg, Michael Tingley. Crowdsourcing with Contextual Uncertainty , ACM SIGKDD 2022
  • Feynman Liang, Nimar Arora, et al. Accelerating Metropolis-Hastings with Lightweight Inference Compilation, AISTATS 2021.
  • Nazanin Khosravani Tehrani, Nimar Arora, et al. Bean Machine: A Declarative Probabilistic Programming Language For Efficient Programmable Inference. The 10th International Conference on Probabilistic Graphical Models. 2020..
  • Ronan Le Bras, Nimar Arora, Noriyuki Kushida, et al. NET-VISA from Cradle to Adulthood. A Machine-Learning Tool for Seismo-Acoustic Automatic Association. Pure Appl. Geophys. (2020).
  • PPLBench: Evaluation Framework For Probabilistic Programming Languages. PROBPROG 2020.
  • Nimar Arora, Nazanin Khosravani Tehrani, Kinjal Divesh Shah, Michael Tingley, Yucen Lily Li, Narjes Torabi, David Noursi, Sepehr Akhavan Masouleh, Eric Lippert, and Erik Meijer. Newtonian Monte Carlo: a second-order gradient method for speeding up MCMC. Advances in Approximate Bayesian Inference 2019 (also in StarAI 2020).
  • Pierrick Mialle, David Brown, and Nimar Arora. Advances in operational processing at the international data centre. In Infrasound Monitoring for Atmospheric Studies, pp. 209-248. Springer, Cham, 2019.
  • Jessica Ai, Nimar S. Arora, Ning Dong, Beliz Gokkaya, Thomas Jiang, Anitha Kubendran, Arun Kumar, Michael Tingley, and Narjes Torabi. HackPPL: a universal probabilistic programming language. In Proceedings of the 3rd ACM SIGPLAN International Workshop on Machine Learning and Programming Languages, pp. 20-28. 2019.
  • Nimar Arora, Stuart Russell, and Erik Sudderth. NET‐VISA: Network processing vertically integrated seismic analysis. Bulletin of the Seismological Society of America 103, no. 2A (2013): 709-729. (Also in AAAI 2011 and NIPS 2010).
  • Nimar Arora, Rodrigo de Salvo Braz, Erik Sudderth, and Stuart Russell. Gibbs sampling in open-universe stochastic languages. In Proceedings of the Twenty-Sixth Conference on Uncertainty in Artificial Intelligence, pp. 30-39. 2010.
  • Lik Wong, Nimar Arora, Lei Gao, Thuvan Hoang, and Jingwei Wu. Oracle streams: A high performance implementation for near real time asynchronous replication. In 2009 IEEE 25th International Conference on Data Engineering, pp. 1363-1374. IEEE, 2009.
  • Nimar Arora, Robert D. Blumofe, and C. Greg Plaxton. Thread scheduling for multiprogrammed multiprocessors. Theory of computing systems 34, no. 2 (2001): 115-144.

    • Selected Patents

  • Sean Lehouillier, Hung V. Tran, Vasanth Rajamani, Nimar S. Arora, and Lik Wong. Dependency-aware transaction batching for data replication. U.S. Patent 10,191,932, issued January 29, 2019.
  • Edwina M. Lu, James W. Stamos, Nimar S. Arora, Lik Wong, Haobo Xu, Thuvan Hoang, Byron Wang, and Lakshminarayanan Chidambaran. Statement-level and procedural-level replication. U.S. Patent 9,569,514, issued February 14, 2017.
  • Lik Wong, Nimar Arora, Lei Gao, Thuvan Hoang, and Haobo Xu. High performant information sharing and replication for single-publisher and multiple-subscriber configuration. U.S. Patent 9,230,002, issued January 5, 2016.
  • Lik Wong, Nimar S. Arora, Anand Lakshminath, Jingwei Wu, Lei Gao, and Thuvan Hoang. Combining capture and apply in a distributed information sharing system. U.S. Patent 8,799,213, issued August 5, 2014.
  • Nimar S. Arora Estimating performance of application based on automatic resizing of shared memory for messaging. U.S. Patent 7,937,257, issued May 3, 2011.
  • Lik Wong, Thuvan Hoang, Nimar Singh Arora, and Jun Yuan. Replicating and sharing data between heterogeneous data systems. U.S. Patent 7,783,601, issued August 24, 2010.
  • James W. Warner, Zhen Hua Liu, Sundeep Abraham, Muralidhar Krishnaprasad, Geeta Arora, Ravi Murthy, Sivasankaran Chandrasekar, Lik Wong, and Nimar S. Arora. Efficient replication of XML data in a relational database management system. U.S. Patent 7,853,573, issued December 14, 2010.
  • Lik Wong, Nimar S. Arora, Cristina Schmidt, Lei Gao, and Thuvan Hoang. Checkpoint-free in log mining for distributed information sharing. U.S. Patent 7,801,852, issued September 21, 2010.
  • Wong, Lik, James William Stamos, and Nimar Singh Arora. On-demand multi-version data dictionary to support distributed applications. U.S. Patent 7,287,034, issued October 23, 2007.
  • Alan Demers, James William Stamos, Lewis S. Kaplan, and Nimar Arora. Method of applying changes to a standby database system. U.S. Patent 6,980,988, issued December 27, 2005.
  • Nimar Arora, Nimar. Incremental refresh of materialized views for many-to-many relationships. U.S. Patent 6,708,179, issued March 16, 2004.