Experience

Academic Research

Deep Intermodal Video Analytics intends to develop robust automated activity detection for a multi-camera streaming video environment.
Working on the system integration and develop the pipeline provided from Kitware to run an end to end object detection and event localization task.
Created a docker image with all software dependencies installed with errors resolved working in conjunction with the Kitware Team.

Designed a crawler to crawl the web and search relevant sound samples for training gunshot audio event detectors.
Performed a literature survey on acoustic analysis of Gunshot recordings and understood how the information could be leveraged to build stronger classifiers.
Achieved a baseline performance of 20 percent mean average precision to detect type of gun from acoustic analysis of gunshots and draw further inferences from audio analysis
Developing a user interface to display gunshots and performing testing on the real world incidents such as Las Vegas Mass Shooting attack to validate the results

Developed a web-based artificial intelligence system to continuously crawl the web, detect and index sounds and automatically learn their meanings, associations, and semantics to events, objects and places with minimal human intervention.
Designed a crawler to continuously crawl the web in search of sound samples for self-training audio detectors.
Created sound vocabularies and performed audio content analysis using machine learning algorithms.
Developed an intuitive web interface to obtain user feedback for semi-supervised learning accessible at nels.cs.cmu.edu.
Improving sound detectors through enhanced database of crawled sounds, implementing feature extraction techniques, classifier fusion and artificial neural networks approach

Developed an in-house verification tool to simultaneously verify 52 subsystems delivered to partners.
First Engineer to verify a chip to chip communication in a multi chip scenario where complete functionality verification task’s responsibility was on my shoulders. Delivered to external partner without potential bugs.
Received Bravo award in 2017 and 2016 for flawless execution across delivery cycles and fixed 5 critical RTL bugs impacting multiple teams across the globe.

Compute subsystem verification.
Write test cases to support Power transitions targeting CPUs, GPUs and other IP blocks in a design
Scalable code for power verification to be used across multiple teams for various projects
Verification methodolgies include C-based testing, development of UVM environment and Formal analysis.
Found and fixed critical bugs in the RTL design.
Received Bravo award for 2x speedup in delivery of tasks

Given an audio test file, we aim to automatically learn meanings, associations, and semantics between different audio clips through a web based intelligence system
Able to categorize sound based on a binary classifier indicating whether sound belongs to a particular class or not

Designed of Low Power Interface for AMBA 4 and ACE Protocol as a member of Systems and Software group.
Given a multi cluster CPU arrangement, synchronization of events from one CPU to another was designed
Deep knowledge of ARMv8 Architecture and flow of design and verification from the description of engineering specification skills were acquired during this period

Contributed to design of a Solar Mobile charger, Charge controller Circuit, MPPT Charge Controller, Wireless modules for solar tracking, and Light Seeking BOT.