Kritti Sharma
Graduate Student Researcher at Caltech- Claim this Profile
Click to upgrade to our gold package
for the full feature experience.
Topline Score
Bio
Credentials
-
Convolutional Neural Networks
CourseraJul, 2020- Oct, 2024 -
Deep Learning Specialization
CourseraJul, 2020- Oct, 2024 -
Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization
CourseraJul, 2020- Oct, 2024 -
Neural Networks and Deep Learning
CourseraJul, 2020- Oct, 2024 -
Sequence Models
CourseraJul, 2020- Oct, 2024 -
Structuring Machine Learning Projects
CourseraJul, 2020- Oct, 2024 -
Blockchain Basics
CourseraJun, 2020- Oct, 2024
Experience
-
Caltech
-
United States
-
Research Services
-
700 & Above Employee
-
Graduate Student Researcher
-
Sep 2022 - Present
The enigmatic short-duration radio pulses, fast radio bursts (FRB), were discovered nearly a decade ago in our dynamic night sky. Still, the underlying astrophysical mechanisms that trigger these transients are poorly understood, owing to various probable progenitor scenarios. The stellar population in the neighborhood of extragalactic transients and their delay-time distributions provide critical insight into their progenitor physics. Historically, such statistical endeavors have played a… Show more The enigmatic short-duration radio pulses, fast radio bursts (FRB), were discovered nearly a decade ago in our dynamic night sky. Still, the underlying astrophysical mechanisms that trigger these transients are poorly understood, owing to various probable progenitor scenarios. The stellar population in the neighborhood of extragalactic transients and their delay-time distributions provide critical insight into their progenitor physics. Historically, such statistical endeavors have played a crucial role in unveiling the nature of progenitors of several transients, such as gamma-ray bursts and supernovae. The venture of characterizing host galaxies of FRBs is essential to disentangle the proposed spectrum of formation channels, ranging from young magnetars formed during core-collapse supernovae to binary neutron star mergers in early-type galaxies. The major challenge in achieving this goal is to obtain their sub-arcsecond localizations. The Deep Synoptic Array-2000 (DSA-2000) aims to bridge this gap by enabling precision localizations with sustained, high-speed monitoring at an FRB discovery rate of 2,000/year. The aim of this project is to study the FRB host galaxies localized by DSA-110, a prototype to test key technologies for DSA-2000, during its commissioning stage by digging up and constraining their star formation histories. Show less The enigmatic short-duration radio pulses, fast radio bursts (FRB), were discovered nearly a decade ago in our dynamic night sky. Still, the underlying astrophysical mechanisms that trigger these transients are poorly understood, owing to various probable progenitor scenarios. The stellar population in the neighborhood of extragalactic transients and their delay-time distributions provide critical insight into their progenitor physics. Historically, such statistical endeavors have played a… Show more The enigmatic short-duration radio pulses, fast radio bursts (FRB), were discovered nearly a decade ago in our dynamic night sky. Still, the underlying astrophysical mechanisms that trigger these transients are poorly understood, owing to various probable progenitor scenarios. The stellar population in the neighborhood of extragalactic transients and their delay-time distributions provide critical insight into their progenitor physics. Historically, such statistical endeavors have played a crucial role in unveiling the nature of progenitors of several transients, such as gamma-ray bursts and supernovae. The venture of characterizing host galaxies of FRBs is essential to disentangle the proposed spectrum of formation channels, ranging from young magnetars formed during core-collapse supernovae to binary neutron star mergers in early-type galaxies. The major challenge in achieving this goal is to obtain their sub-arcsecond localizations. The Deep Synoptic Array-2000 (DSA-2000) aims to bridge this gap by enabling precision localizations with sustained, high-speed monitoring at an FRB discovery rate of 2,000/year. The aim of this project is to study the FRB host galaxies localized by DSA-110, a prototype to test key technologies for DSA-2000, during its commissioning stage by digging up and constraining their star formation histories. Show less
-
-
-
-
Research Student
-
Jul 2021 - Aug 2022
Hierarchical black hole mergers are expected to be one of the most significant contributors to Intermediate Mass Black Hole (IMBH) binary mergers. This formation channel can produce binaries with a large amount of mass asymmetry, and therefore, their mergers can produce gravitational waves with a measurable amount of higher-order mode content. The search for IMBH binaries poses a unique challenge. These signals are expected to be highly short-lived (often lasting less than one-tenth of a… Show more Hierarchical black hole mergers are expected to be one of the most significant contributors to Intermediate Mass Black Hole (IMBH) binary mergers. This formation channel can produce binaries with a large amount of mass asymmetry, and therefore, their mergers can produce gravitational waves with a measurable amount of higher-order mode content. The search for IMBH binaries poses a unique challenge. These signals are expected to be highly short-lived (often lasting less than one-tenth of a second) within the current detector bandwidth. Therefore, they can be confused with the frequently occurring noisy glitches whose morphology they often resemble. Moreover, in highly asymmetric systems, significant waveform modulations due to higher-order modes make it difficult to model and detect these signals buried in the detector noise. The time-frequency representation of these signals has distinctive features owing to these waveform modulations. I have developed a Deep Transfer Learning algorithm, THAMES, which extracts such characteristic features from the spectrograms to distinguish between IMBH binary signals and several classes of glitches. This algorithm looks for time coincident triggers in two detectors and ranks the coincident events using a novel detection statistic. I have evaluated the sensitivity of our algorithm with asymmetric, nearly edge-on IMBH binary simulations in Gravitational Wave open data. This algorithm notably outperforms the optimized template-based PyCBC search for signals with higher-order mode content. Show less Hierarchical black hole mergers are expected to be one of the most significant contributors to Intermediate Mass Black Hole (IMBH) binary mergers. This formation channel can produce binaries with a large amount of mass asymmetry, and therefore, their mergers can produce gravitational waves with a measurable amount of higher-order mode content. The search for IMBH binaries poses a unique challenge. These signals are expected to be highly short-lived (often lasting less than one-tenth of a… Show more Hierarchical black hole mergers are expected to be one of the most significant contributors to Intermediate Mass Black Hole (IMBH) binary mergers. This formation channel can produce binaries with a large amount of mass asymmetry, and therefore, their mergers can produce gravitational waves with a measurable amount of higher-order mode content. The search for IMBH binaries poses a unique challenge. These signals are expected to be highly short-lived (often lasting less than one-tenth of a second) within the current detector bandwidth. Therefore, they can be confused with the frequently occurring noisy glitches whose morphology they often resemble. Moreover, in highly asymmetric systems, significant waveform modulations due to higher-order modes make it difficult to model and detect these signals buried in the detector noise. The time-frequency representation of these signals has distinctive features owing to these waveform modulations. I have developed a Deep Transfer Learning algorithm, THAMES, which extracts such characteristic features from the spectrograms to distinguish between IMBH binary signals and several classes of glitches. This algorithm looks for time coincident triggers in two detectors and ranks the coincident events using a novel detection statistic. I have evaluated the sensitivity of our algorithm with asymmetric, nearly edge-on IMBH binary simulations in Gravitational Wave open data. This algorithm notably outperforms the optimized template-based PyCBC search for signals with higher-order mode content. Show less
-
-
-
Space Technology and Astrophysics Research (STAR) Lab, IIT Bombay
-
Mumbai, Maharashtra, India
-
Research Student
-
Aug 2020 - Dec 2021
I led the automation of non-sidereal observations data acquisition and operation for the 70 cm fully robotic GROWTH-India Telescope. I also automated the non-sidereal data reduction for GROWTH-India Telescope, where we developed, tested and validated our robust and modular data reduction pipeline - Astreaks. I have also been involved in the discovery of Near-Earth Asteroids (NEAs) using data from the Zwicky Transient Facility, and follow-up of NEAs and outbursting comets using the GROWTH-India… Show more I led the automation of non-sidereal observations data acquisition and operation for the 70 cm fully robotic GROWTH-India Telescope. I also automated the non-sidereal data reduction for GROWTH-India Telescope, where we developed, tested and validated our robust and modular data reduction pipeline - Astreaks. I have also been involved in the discovery of Near-Earth Asteroids (NEAs) using data from the Zwicky Transient Facility, and follow-up of NEAs and outbursting comets using the GROWTH-India Telescope. I co-discovered the then known closest asteroid fly-by without impact, 2020 QG. I've also studied the episodically active asteroid (6478) Gault, active asteroid 2005 XR132, first-ever known inter-Venusian asteroid 2020 AV2 and main belt active asteroid 2005 QN173 with GROWTH Collaboration. Show less I led the automation of non-sidereal observations data acquisition and operation for the 70 cm fully robotic GROWTH-India Telescope. I also automated the non-sidereal data reduction for GROWTH-India Telescope, where we developed, tested and validated our robust and modular data reduction pipeline - Astreaks. I have also been involved in the discovery of Near-Earth Asteroids (NEAs) using data from the Zwicky Transient Facility, and follow-up of NEAs and outbursting comets using the GROWTH-India… Show more I led the automation of non-sidereal observations data acquisition and operation for the 70 cm fully robotic GROWTH-India Telescope. I also automated the non-sidereal data reduction for GROWTH-India Telescope, where we developed, tested and validated our robust and modular data reduction pipeline - Astreaks. I have also been involved in the discovery of Near-Earth Asteroids (NEAs) using data from the Zwicky Transient Facility, and follow-up of NEAs and outbursting comets using the GROWTH-India Telescope. I co-discovered the then known closest asteroid fly-by without impact, 2020 QG. I've also studied the episodically active asteroid (6478) Gault, active asteroid 2005 XR132, first-ever known inter-Venusian asteroid 2020 AV2 and main belt active asteroid 2005 QN173 with GROWTH Collaboration. Show less
-
-
-
Goldman Sachs
-
United States
-
Financial Services
-
700 & Above Employee
-
Summer Analyst
-
Jun 2021 - Jul 2021
I aimed at the art of machine learning programming and undertook the empirical adventure of building frameworks to control rebill rates and identify bogus claims in disputed transactions data, closely working with transactions, fraud capabilities, disputes strategy, investigation specialists and modelling teams. I devised metrics for driver and target population analysis and queried transactions and disputes data to compute metrics using SQL and PySpark in databricks. I examined these metrics… Show more I aimed at the art of machine learning programming and undertook the empirical adventure of building frameworks to control rebill rates and identify bogus claims in disputed transactions data, closely working with transactions, fraud capabilities, disputes strategy, investigation specialists and modelling teams. I devised metrics for driver and target population analysis and queried transactions and disputes data to compute metrics using SQL and PySpark in databricks. I examined these metrics for rationality using their graphical representation. I ideated two hypothesis, pertaining to the problem statement and investigated their validity using data, where one hypothesis was substantiated and one was rejected after laborious consideration of statistical inferences. I deployed a decision tree classifier with XG-boost to train a machine learning model using the evaluated metrics, identified top features and certified model performance on out-of-time data using the modular program developed. I deliberated with my ingenious managers and field experts over the efficacy of results from the model developed, model integration into real-time pipeline execution, correlated business outcomes and long-term impact of the project. I got the opportunity to articulate and present my models to the leadership, including the motivation of my problem statement, analysis framework, modelling framework, observations, results from the synchronised implementation of the model in ongoing disputes and business outcomes. Show less I aimed at the art of machine learning programming and undertook the empirical adventure of building frameworks to control rebill rates and identify bogus claims in disputed transactions data, closely working with transactions, fraud capabilities, disputes strategy, investigation specialists and modelling teams. I devised metrics for driver and target population analysis and queried transactions and disputes data to compute metrics using SQL and PySpark in databricks. I examined these metrics… Show more I aimed at the art of machine learning programming and undertook the empirical adventure of building frameworks to control rebill rates and identify bogus claims in disputed transactions data, closely working with transactions, fraud capabilities, disputes strategy, investigation specialists and modelling teams. I devised metrics for driver and target population analysis and queried transactions and disputes data to compute metrics using SQL and PySpark in databricks. I examined these metrics for rationality using their graphical representation. I ideated two hypothesis, pertaining to the problem statement and investigated their validity using data, where one hypothesis was substantiated and one was rejected after laborious consideration of statistical inferences. I deployed a decision tree classifier with XG-boost to train a machine learning model using the evaluated metrics, identified top features and certified model performance on out-of-time data using the modular program developed. I deliberated with my ingenious managers and field experts over the efficacy of results from the model developed, model integration into real-time pipeline execution, correlated business outcomes and long-term impact of the project. I got the opportunity to articulate and present my models to the leadership, including the motivation of my problem statement, analysis framework, modelling framework, observations, results from the synchronised implementation of the model in ongoing disputes and business outcomes. Show less
-
-
-
Indian Institute of Technology, Bombay
-
India
-
Higher Education
-
700 & Above Employee
-
Teaching Assistant
-
Jan 2021 - Apr 2021
Guided a batch of 30+ Students in the selection of course project topics and estimated their observation requirements, thus helping them gain an experience in submitting observation proposals for 0.7 m GROWTH-India Telescope. Scheduled observations for all teams, acquired data, did basic processing (bias correction, flat fielding, cosmic ray removal and photometric calibration) and taught data analysis to students using basic astronomy tools like DS9 and APT. Guided a batch of 30+ Students in the selection of course project topics and estimated their observation requirements, thus helping them gain an experience in submitting observation proposals for 0.7 m GROWTH-India Telescope. Scheduled observations for all teams, acquired data, did basic processing (bias correction, flat fielding, cosmic ray removal and photometric calibration) and taught data analysis to students using basic astronomy tools like DS9 and APT.
-
-
-
-
Manager
-
Apr 2020 - Apr 2021
Leading a team of 7 conveners and 3 volunteers to foster enthusiasm in Astronomy, tending to a community of about 100 in the institute and an online community of more than 5,500 enthusiasts. Ideating the LIGO Student Reading Group with the aim of bringing together the knowledge of gravitational wave theory, astronomical source modeling, data analysis, computational strategy and precision instrumentation. Supervised Krittika Summer Projects, which comprised of 2 week long daily tutorial sessions… Show more Leading a team of 7 conveners and 3 volunteers to foster enthusiasm in Astronomy, tending to a community of about 100 in the institute and an online community of more than 5,500 enthusiasts. Ideating the LIGO Student Reading Group with the aim of bringing together the knowledge of gravitational wave theory, astronomical source modeling, data analysis, computational strategy and precision instrumentation. Supervised Krittika Summer Projects, which comprised of 2 week long daily tutorial sessions on Numerical and Scientific Computing catering to an online community of about 2,400 along with 6 projects guided by experienced mentors, with an excellent response from multiple IITs and other colleges in India.
-
-
Convener
-
Apr 2019 - Mar 2020
As part of a team of 8, was responsible for organizing several institute-wide events such as lectures, workshops, group discussions, documentary screenings and organizing trips to various observatories. Sensitized people to astrophotography by conducting night sky observation sessions and familiarized beginners in astronomy with basic concepts of telescope handling and observation planning. Organized competitions such as Astromania(the annual institute astronomy quiz), Scientific Computation… Show more As part of a team of 8, was responsible for organizing several institute-wide events such as lectures, workshops, group discussions, documentary screenings and organizing trips to various observatories. Sensitized people to astrophotography by conducting night sky observation sessions and familiarized beginners in astronomy with basic concepts of telescope handling and observation planning. Organized competitions such as Astromania(the annual institute astronomy quiz), Scientific Computation General Championship and Observation Planning General Championship for astronomy amateurs in the campus.
-
-
-
-
Course Moderator
-
Jun 2020 - Jul 2020
As part of a team of 5, mentored a batch of around 250 in generating Computational Models with Python to aid in solving scientific problems like gravity simulator, heat transfer, predator-prey model, epidemiology, etc. Designed course material for learning how to solve Matrix and Differential Equations computationally, including Numerical Methods such as Euler’s method, Finite Difference method and Runge-Kutta algorithm.
-
-
Project Mentor
-
Apr 2020 - Jun 2020
Mentored a group of 3 students, guiding them for a three month long Reading Project, by sharing resources and research material on General and Observational Astronomy and fostering active discussions within the group. Topics covered included space exploration, missions of ISRO and NASA, ISS, Hubble telescope, paradoxes in cosmology, galaxies, hydrostatic equilibrium of stars, stellar energy sources and supernova explosions.
-
-
Education
-
2022 - 2027
Caltech
Doctor of Philosophy - PhD, Astrophysics -
2018 - 2022
Indian Institute of Technology, Bombay
Bachelor of Technology - BTech with Honors, Mechanical Engineering
Community
