M.S. in Computer Science
University of California San Diego
Jun 2019
La Jolla, CA
  • GPA: 3.96/4.00
  • Specialization: Artificial Intelligence
B.Engg. (honors) in Electrical and Electronics Engineering
Birla Institute of Technology and Science (BITS Pilani)
Jun 2015
Pilani, India
  • GPA: 9.23/10.00
  • Graduated with distinction and among top 5%
Online Courses
  • Self Driving Car Nanodegree, Udacity
  • Artificial Intelligence Nanodegree, Udacity
  • Data Science at Scale Specialization, University of Washington, Coursera

Work Experience

Machine Learning Engineer II
Twitter Inc.
Aug 2019 - Present
San Francisco, CA
  • Working with MLX team for unification of Recommender Systems across Twitter
Graduate Student Researcher
UC San Diego
Sep 2017 - Jun 2019
La Jolla, CA
  • Created module to detect and track pedestrians and vehicles. Detection is done using Yolov3 implemented in pytorch and tracking is done using Kalman Filter
  • Created deep learning module for free space detection (semantic image segmentation) using Fully Connected Network in Pytorch
Twitter Inc.
Jun 2018 - Sep 2018
San Francisco, CA
  • Wrote data pipeline, using Scalding and PySpark, to obtain high-quality muti-class multi-labelled tweet datasets
  • Trained bi-LSTM models, using Tensorflow's Estimator API, to obtain tweet embeddings by minimizing cross entropy loss for multi-class tweet classification
  • Studied effectiveness and transfer-ability of the model by using the learned embedding for different downstream tasks
Data Scientist
Media IQ Digital
Feb 2016 - Aug 2017
Bengaluru, India
  • Implemented a fast and scalable MapReduce Framework (Hadoop, Java) leveraging boolean logic and cache to perform complex pattern matching obtaining over 120x speedup
  • Developed and deployed a fault tolerant behavioral ad-targeting pipeline to handle over 50TB of data and obtained 10-15% uplift in many campaigns
  • Developed a proof-of-concept of a scalable architecture for a Client-Side Behavioral User Profiling for personalized ad-targeting thereby restricting exchange of user's browsing data with ad-servers
Application Developer
Jun 2015 - Jan 2016
Bengaluru, India
  • Created an SQL utility (precision: 88%) for cleansing geographical datasets to get rid of typographical errors and duplicate records using phonetic algorithm like Soundex and NYSIIS
  • Worked on development of an MVVC framework to reduce server load and fetch and display business reports dynamically by runtime modification of HTML nodes
Mitacs Research Intern
University of Manitoba
May 2014 - Aug 2014
Winnipeg, Canada
  • Worked on development of a 3D Game Engine in C++ for Oculus Rift for spatial capability assessment of humans
  • Created modules in OpenCV (Python) for recreation of virtual 3d-world from the output obtained from a Laser-based SLAM


Lessons learned from deploying autonomous vehicles at UC San Diego
Field and Service Robotics (Tokyo, JP, August 2019)
  • Discussed brief overview of the overall design and the design decisions for construction of vehicles for last-mile delivery
  • Discussed design and challenges of vehicles for the micro-mobility challenge based on open source Autoware system
  • Proposed requirements for roboust systems that include a robust control design, a shift towards increased use of image data over LiDAR data, handling of a richer set of vehicles / pedestrians in a last mile scenario, and overall system characterization and evaluation
Transfer of Expertise in Deep Neural Networks
VSS, Tampa FL, May 2019
  • Studied the attention map of expert and novice birdwatchers by modeling them as deep neural network (VGG-16) (experts is defined as systems that categorize stimuli at a subordinate level, and novices are defined as systems that categorize the same stimuli at a coarser grain)
  • The attention map used by the expert network has higher entropy, and smaller, local features than the novice networks (suggesting that the expert looks at multiple locations to make a classification decision)
  • The attention map used by the expert can be used to train the novice, resulting in faster training and better performance. This to be analogous to the expert telling the novice where to look for discriminative regions of the image
Applications of Deep Learning to Autonomous Vehicles
International Conference on Business Analytics and Intelligence, IIM Bangalore, Dec. 2017
  • Explored some of the feature selection and preprocessing techniques to make deep learning models robust
  • Presented a Deep Learning based approach to perform various tasks like traffic sign classification, object detection, semantic segmentation and lane detection
  • Explored the techniques in Deep Learning that are currently being used in the industry for the successful navigation of autonomous vehicles
Electronics Aid for Elder and Sick
Recent Advances and Innovations in Engineering, IEEE, Poornima University, Jaipur, May 2014
  • Proposed a wireless switch board for easy-accessibility in an hospital-like environment for controlling various electrical appliances like light, fan, etc. without interference with similar wireless switches in adjoining room
  • The data signal is communicated to receiver wirelessly by IR Remote Controller. The receiver consists of IR Receiver, TSOP1838, to read data signal operating at 38 kHz and a microcontroller to process data and communicate trigger signal to change the state of switch


  • Techniques: Deep Learning, Machine Learning, Data Science, Computer Vision, NLP
  • Languages: Python, C/C++, Java, Matlab, HTML, CSS, Javascript
  • Tools/Packages: ROS, tensorflow, keras, pytorch, scikit-learn, openCV, pyspark, hadoop, hive, pandas, numpy, Android SDK