top of page

CURRICULUM VITAE

PDFs: Resume

EDUCATION

PROFESSIONAL EXPERIENCE

ELEMENT ENERGY  -  Menlo Park, CA, January 2023 - present

  • Leading the development and deployment of deep convolutional network (CNN)-based platform to predict states of degradation (loss of lithium, loss of active material) and remaining useful life (RUL) in large-scale fleet (>200,000) of Lithium-ion batteries with fractional errors of 0.01 – 0.03.

  • This novel solution uses dynamic time warping to convert time series curves of batteries and their pristine counterparts to two dimensional images and analyze them with artificial intelligence platforms based on EfficientNet based transfer learning and CNNs.

  • Used Variational Autoencoders, Mann-Kendall Trend Test and Thiel Sen estimators to determine trends and robustly detect outliers in degradation metrics of in batteries with recall scores >90% and F1 scores >95%.

​

DEEPFENCE INC.  -  Palo Alto, CA, February 2021 - January 2023

  • Built innovative Go based anomaly detection pipeline which clusters network data from traffic between and across virtual machines (VM), K8s pods, containers, processes, and external hosts, and uses inter quartile range (IQR) to detect anomalies at runtime.

  • The aforesaid anomaly detection engine detects malicious traffic in north-south and east-west connections in environments with over 10,000 nodes and over 60,000 K8s pods with high true positive rates, low false positive rates in class imbalanced environments.

  •  This innovative method includes features for concurrent consideration of both categorical and numerical data, smart initialization of cluster centers and automatic detection of most appropriate number of clusters.

  • The above-mentioned Go based anomaly detection pipeline improved detection times of anomalies from 5 minutes to 20 seconds. Also, it delivered over 1.5 times the accuracies of k-prototype, k-means and k-modes clustering based anomaly detection methods.

​

​

CISCO SYSTEMS INC.   -  San Jose, CA, September 2019 - January 2021​

  • Developed core components of self-healing networks which learn from telemetry data from six continents to predict user experience issues in various forecast horizons and provide problem solving options.

  • The afore-mentioned predictive technology helps in improving reliability and performance of networks in over 50 countries.

  • Designed machine learning algorithm for peer anomaly detection in time series data using telemetry data from over 100 customers across 50 countries.

ACADEMIC RESEARCH EXPERIENCE

Semi Supervised Learning for Diabetic Retinopathy

Machine Learning Department, CMU Fall 2018

  • Built a semi-supervised deep learning architecture for Diabetic Retinopathy (DR) detection from retinal fundus images in dataset put together by three hospitals in France.

  • Used Canny edge detection to preprocess the retinal fundus images to detect important features like veins, and lesions like exudates, microaneurysms, and hemorrhages.

  • This convolutional auto-encoder based pipeline for DR detection, achieves 2% improvement over the ResNet18 baseline on medical image datasets, viz, Messidor and Indian Diabetic Retinopathy dataset.

​

Recognition of Electronic Components in Printed Circuit Boards

Machine Learning Department, CMU Fall 2018

  • Used genes, protein and DNA data from 33 labs and research publications with 17,000 cell expression samples.

  • Designed multi-stage machine learning pipeline based on principal component analysis to accurately predict cell types.

  • Reduced the number of features from ~80,000 to ~350 and achieved 94% accuracy in determining cell types.

​

Prediction of Cell Types from Gene Data

Machine Learning Department, CMU Fall 2018

  • Used genes, protein and DNA data from 33 labs and research publications with 17,000 cell expression samples.

  • Designed multi-stage machine learning pipeline based on principal component analysis to accurately predict cell types.

  • Reduced the number of features from ~80,000 to ~350 and achieved 94% accuracy in determining cell types.

​

Attribute Prediction in Networks

Machine Learning Department, CMU Spring 2017

  • Developed predictor (Proclivity Propagation) to impute missing attribute values and detect outliers in friendship networks.

  • Proclivity Propagation is capable of capturing homophily, heterophily, self and cross proclivities in attributed networks.

  • This novel predictor is unique in the sense that it gives confidence intervals on accuracies that it predicts.

  • Used a state-of-the-art network correlation matrix called PROclivity index for attributed Networks (PRONE) to find attributes relevant for prediction of missing attribute values.

  • Tests of Proclivity Propagation on Facebook100 dataset give prediction accuracies as high as 85%, 78%, 75% and 69% for attributes like year, dormitory, status and gender.

  • This new method gives better prediction accuracies than standard machine learning and statistical techniques like Support Vector Machine classifiers and Low Rank Matrix Completion.

​

Growth rate of the Universe

McWilliams Center for Cosmology, CMU Fall 2015 – Spring 2016

  • Led project about measurement of correlation of 1.2 million galaxies in largest ever 3D map of the Universe obtained from the Sloan Digital Sky Survey Telescope (MNRAS 469, 1369).

  • Designed novel estimator for measuring correlations in 1.2 million galaxies. Detected values of cosmological parameters like Hubble constant with very high precision (~3%).

  • Used C++ to compute pair queries and wrapped the C++ library as a Python extension module.

​

Facial Action Unit Detection

Dept. of Electrical and Computer Enginering, CMU Spring 2015

  • Automatic recognition of facial expressions and emotions using 46 unique action units around eyes and lips.

  • Built pipeline to implement Hierarchical Support Vector Machine classifier to classify and detect facial expressions in videos of human faces (https://www.youtube.com/watch?v=-KzRn1RBRbw).

  • Achieved accuracies of 95%, 95% and 75% during detection of facial expressions like open mouth, neutral face and smile.

​

Teaching Assistant

Carnegie Mellon University 2014-2018

•  Five years of teaching experience across two departments (Physics, Mathematics) in Carnegie Mellon University.

•  Taught nine full-credit courses in Physics Department, CMU and one course in Mathematics Department, CMU; average ratings of 4.9 out of 5.0.

 

TECHNICAL EXPERTISE

  • LANGUAGES : Python, C++, Cython, Go, SQL, Mathematica, MATLAB              •   PLATFORMS:      AWS, Kubernetes (K8s), Docker

  • LIBRARIES :  PyTorch, Tensorflow, Keras, OpenCV, Scikit-learn                               •   FRAMEWORKS:  Spark 

SELECTED PUBLICATIONS

COMMUNITY INVOLVEMENT

AWARDS AND RECOGNITION

  • The Hugh D. Young Graduate Student Teaching Award, Mellon College Science (April 2017): The award is given to one graduate student in the Mellon College of Science each year in recognition of effective teaching.

  • The Physics Department Teaching Assistant Award, CMU (Nov 2015): This award is given annually to the physics graduate student who best exemplifies the high standards of education in the Physics Department by going beyond the normal expectations for a Teaching Assistant.

  • The Department Research Fellowship, Physics Dept., CMU (2013-2014): This award is given annually to chosen students who demonstrate outstanding aptitude for research.

  • Gold Medals for Best Academic Performance, National Institute of Science Education and Research, India (2012, 2011, 2010, 2009): This award is given annually to the best student of the university.

  • Best Thesis Award, National Institute of Science Education and Research, India (2012): This award is given to the best undergraduate thesis in the graduating batch of students.

  • Sarat Chandra-Annapurna Award for Outstanding Academic Performance, National Institute of Science Education and Research, India (2012): This award is given for excellence in academics in the university.

  • Visiting Students Research Program Fellowship, Tata Institute of Fundamental Research (2010): This fellowship is given to meritorious undergraduate students for research in science and technology.

  • Indian Academy of Sciences Research Fellowship, Indian Academy of Sciences (2009): This fellowship is given to meritorious undergraduate students for research in science and technology.

  • National Initiative in Undergraduate Sciences Fellowship, Homi Bhabha Center for Science Education (2008): This fellowship is given to meritorious undergraduate students for research in science and technology.

  • The Innovation in Science Pursuit for Inspired Research Scholarship, Dept. of Science and Technology, Govt. of India (2007): This scholarship is given to meritorious undergraduate students for excellence in science and technology.

  • Ranked first in National Entrance Screening Test, Govt. of India, (2007): 1st out of approximately 100,000 applicants.

  • Kishore Vaigyanik Protsahan Yojana (Young Scientist Encouragement Program Fellowship), Dept. of Science and Technology, Govt. of India (2006): This award is given to exceptionally motivated students for pursuing research career in science and technology.

  • Mathematics Olympiad, National Board of Higher Mathematics (2003, 2004, 2004, 2006)

VOLUNTEER WORK AND ORGANIZATIONS

  • Vice President of Finance Committee, Graduate Student Assembly, CMU (April 2017 – Sept 2017)

  • Representative of Dept. of Physics to Graduate Student Assembly, CMU (Feb 2015 – April 2017)

  • Member of GuSH-Crosswalk Seed Grant Evaluation Committee, CMU (2015 – 2019)

  • Member of Graduate Student Activities Committee, Mellon College of Science, CMU (2015 – 2019)

  • Student representative to Undergraduate Committee of the Institute, NISER (2009 – 2012)

REFERENCES

Available upon request.

bottom of page