Role: Software Engineering Intern
-
Morty - The intelligent Bot:
Designed and developed a bot which answered queries regarding release builds and their metrics.
Tech used: Azure Bot Framework, Azure Language Understanding Service and Azure Cloud AI, ReactJS, C#.
-
Hackathon:
Won the annual Machine Learning and Data Analytics 2018 hackathon with over 100 participants.
Role: Member of Technical Staff 2
-
Storage Policy Based Management:
Designed and implemented the frontend for VMware neXt generation Vasa Provider (xVP) in Java. This helped
create
performant and scalable VP implementation thus reducing VMware's dependency on partner implementations.
-
Patent:
"Early Risk Detection System" helps detect system states which can lead to a crash and pre-emptively makes
changes to migrate to
a safe state for systems running within VSphere.
-
PowerCLI:
Developed vSAN management commandlets like vsan-health, vsan disk etc for the 2016 release.
-
Hackathons:
Developed Kubernetes management framework on vCloudAir, container
management on multi-cloud and host-based SPBM. Have won Best Project Award (out of 40 other new college
graduate projects) and best market potential award for the above projects.
Other Experience
-
Google Summer of Code (Student Developer):
Developed full stack implementations of microblogs, user profiles and communication amongst the users using
Java for open source organisation Freenet project.
-
TagAlong, MIT Global Startup labs (Developer):
Co-founded a dynamic ride sharing startup which was incubated at MIT startup
labs. Designed and developed the Android application for TagAlong.
-
PES Entrepreneurship cell (Club co-founder, Secretary):
Co-founded and led PES E-Cell. With the help of the amazing
E-Cell team we organised over 20 events and one mega event boasting of 200+ participants raising over INR
300,000
from sponsors like Reliance and Github.
-
Juniper Networks Intern, Application Development:
Conceptualized and built Android application for managing approval related tasks. Used ssl certificates to
provide login rather than the simpler text based username and password verification.
Projects and Publications
-
Understudy Approach - A Multi-Agent
Reinforcement Learning technique:
In this approach we trained agents to learn complex tasks
by combining the models of agents trained on simpler objectives. Applied this approach to encourage
collaboration between two agents whose
objective was to collect multiple coins in a room as efficiently as possible.
-
Super Resolve Videos
using SRGANs:
One of the state of the art techniques at the time of writing, to super resolve an image involves the SRGAN
technique. We investigated the effects extending the SRGANs on videos to super resolve them. We further
modified the loss functions to remove
flickering effects.
-
Publication: An Early Risk Detection and Management System for the Cloud with Log Parser - Computers
in Industry, Elsevier May 2018 https://doi.org/10.1016/j.compind.2018.01.018:
Built a system that combined association rule learning and machine learning to
provide insights on the operations that can put a system in peril, and recommend suitable steps to reduce or
eliminate the risks causing a crash.
-
Improving query latency for Greenplum
Database:
Introduced the concept of user priorities for jobs in queue. These priorities were used to throttle the
compute resources for the inter-connected processes in the distributed database, significantly improving
the response latencies.
Achieved an 80% decrease in response time for high priority users and a 10% decrease in response time for
lower priority users. This project was adjudged the best out of 70 projects under “Technology Impact” in the
Class of 2015.