Recent work
I have developed artifical neural network based classifier in python to classify a given dataset.
Project is available on my github repo
Repository contains sample ANN trained on MNIST dataset that achieves 98.33% test accuracy.
Past work
Experience
I worked as a PGET at Mahindra Susten for a year as an embedded system engineer.
I was resoonsible for designing and developing embedded system for special projects devision at Mahindra susten.
I have designed and developed prototype for VI-logger, 3-phase BLDC controller and other subsystems at Mahindra Susten.
System administrator in PCLAB dept. of electrical engineering, IIT Bombay
Projects
I have developed neural network based classifier in R and Weka to classify sensor measurements as faulty and correct using statistical parameters as the feature set for network.
Only amplitude domain faults offset, random noise and spikes are classified as faults.
Model achieved accuracy of about 98% on test data.
Generalization of the model achieved about 99% accuracy on correlated sensor and about 93% on non-correlated sensor.
I have developed following components and connected them to allow collection and access of data for research in IIT Bombay.
Beaglebone black based gateway for pushing data from ZigBee PAN to server over Ethernet.
Ubuntu based server system using Dell edge T20 and used LAMP stack for collection and access data over PHP REST apis.
Website and Android app for realtime visualization of data and monitoring sensor network.
We have developed prototype (in a team of 3 including myself) ceiling fan regulator that can controlled using Andriod over Bluetooth connection.
Android app features set a sleep duration function so that fan speed can follow body temperature profile of an average healthy human being.
We have developed prototype (in a team of 3 including myself) Intel Edison based prototype for mosquito density sensing alarm.
Mosquito density was sensed using microphones.
It was presented and won first prize in VLSID conference in Banglore Jan 2015
We have developed prototype (in a team of 3 including myself) AVR microcontroller based GLCD-touchscreen hardware to place orders in restaurant.
Three prototypes were developed, 2 acting as customer modules and 1 as a chef module.