Teaching Assistant
The Graduate School at Penn State August 2019 - Present
Develope operating system assignments for undergraduate students and write auto grading scripts.
Manage the team of 11 TAs and LAs to help 400 students in operating systems course.
Data Engineer Intern
KOGENTiX Inc. (now acquired by Accenture) June 2017 - August 2017
Worked on implementing a pipeline architecture for data processing by applying data transformation
for analysis. Applied MapReduce model to significantly improve data processing time. Used HQL and SparkSQL
for observing data.
Research Assistant
Undergraduate Deep Learning Lab (UDLL) June 2018 - May 2019
Collaborated with Astronomy researchers to build deep learning model for detecting defects
caused by alpha particles in radio based telescope images.
Wrote deep neural network model to predict number of defects in an image to an accuracy of 78%.
Data consists of 60% images with 0 or 1-pixel defect in an image making pixel prediction challenging.
Undergraduate Research Assistant
D.A.T.A. Labs August 2017 - December 2019
Improved a Child Safety Game to help low income/unprivileged parents learn how to keep
homes safe for children.
Evolved an android mobile application where users can capture heartbeats, blood pressure, etc.
from live video.
Developed end-to-end website for users to capture vitals and download vital data (https://www.videovitals.org)
Django, React Native, Python, Javascript
- Developed an application that allows user to locate nearby friends close to their location and
to get information about nearby social events or group meetings.
- Managed back-end team responsible for creating and updating database with user data and
developing secure and fast servers
- Expanded user-interface with functionalities such as messaging, creating groups and events,
poking a friend, and optional real-time locations of users.
- Collaborated with Civil Engineers on landslide detection project using remotely sensed images.
- Applied data augmentations to reduce class imbalance from dataset that increased performance metric by 5%.
- Working on increasing resolution of Digital Elevation Maps data for better pixel mapping with
corresponding landslide image.
C++, Linux, Synchronization, Caching, Google Remote Procedure Call (gRPC)
- Implemented PFS to allow multiple clients to do file operations (open, read, write, seek, close)
on files that are stripped across multiple file servers with no shared physical memory and disk storage.
- Designed a client-side cache using invalidation based protocol to further improve PFS performance.
- Constructed a Distributed Metadata Manager to store and maintain all metadata information associated with
a file.
- Developed hidden Markov model for part of speech (POS) tagging using brown corpus as training data.
- Optimized model to achieve 94% accuracy.
- Derived theoretical proof to show subspace clustering is better than K-means clustering for
high dimensional data.
- Provided theoretical guarantees to make spectral based subspace clustering more robust in
presence of noise and error.
- It has various applications such as image processing, face clustering, motion segmentation, etc.
