Engineering Electrical
Resume:
JESHWANTH CHALLAGUNDLA
Arlington, TX-76010
acfh0n@r.postjobfree.com
Education:
MS-Thesis (Electrical Engineering) University of Texas at Arlington, Arlington, Texas (Aug 13-May 15)
GPA: 4.0/4.0
Thesis topic: An improved second order Deep Learning algorithm.
Working as a lab intern in Image Processing and Neural Networks Lab in University of Texas at Arlington.
Received the IPNN lab scholarship for the entire period of my Masters.
Bachelor of Technology (Electrical and Electronics Engineering) (Aug 09-May 13)
Pondicherry Engineering College, Pondicherry, India. GPA: 8.4/10
SKILLS:
Experienced in developing second order Deep learning algorithms for feature generation, Neural Network
classifiers for Computer Vision and Data Mining applications, Pattern recognition and feature extraction
techniques.
Embedded system design, microcontroller programming, hardware design, simulation and prototyping.
Good Programming skills: C, C++, Matlab, Python, R.
Software: Matlab, MPLAB, Lab View, Ms Office.
Thesis:
Implementing an improved Second order deep learning algorithm
Developed a sparse autoencoder in which output weight optimization is done using Orthogonal Least
Squares and input weights are updated bysteepest descent.
Used Newtons method in order to calculate Multiple Optimal Learning Factor (MOLF) and Optimal Input
Gain (OIG) from Hessians of the outputs.
Gradients are modified using Hidden weight optimization technique
Experimented on various combinations of this three techniques to come up with optimal input weight
updater i.e. the one that creates maximum error decrease per multiply
Training the autoencoders using data from random distribution and stacking them up to form a deep
network and then fine tuning the final network in supervised way using relevant data to form an efficient
feature extractor.
IPNN Lab Projects:
License Plate Recognition (Industry Sponsored Project in IPNNL Lab)
Character segmentation in License plate images using Adaptive Thresholding method
Extracting 2D-DFT features of these segments and classifying these segments into their respective
alphanumeric character using a Neural Network classifier with one hidden layer which is trained using
second order information and output reset.
Used softmax in order to find the segment that has highest probability of having a character and
idealizing its dimension to rest of the segments. This helps to get rid of imperfections in segmentation
Academic Projects:
Timing Intensive Asynchronous Communications Interface based on the DMX512-A Protocol with EF1 Topology
The PC transmitter will accept commands from a PC via an RS-232 interface and will continuously transmit
a serial stream to control up to 512 devices on a RS-485 communication bus. The PC receiver will forward
data received from devices on a communications bus and send these to the PC with the RS-232 interface.
Devices on the bus will extract information out of the asynchronous data stream and will control one or
more devices. They will also send an acknowledgement to the controller when requested.
Designed and tested the simulation model of Digital Communication
Simulated in MATLAB 4 PSK and 4 DPSK modulation techniques. Random symbols were generated and
modulated using 4 PSK and 4 DPSK modulation techniques. These symbols were detected using MAP and
ML detection techniques. Prior to detection, channel equalization was done to account for errors due to
ISI.
Estimated their performance in terms of symbol error probability
MIMO Simulation Study
Simulated and studied the behavior of MIMO systems under several channel conditions and coding
schemes and the deduce quality of communication in each case by spatial multiplexing in the presence of
polarization diversity.
Relevant courses:
Neural Networks- Design of higher order neural network algorithms, piecewise-linear nets, radial basis
function nets and counter-propagation nets.
Statistical Pattern Recognition- Different feature extraction techniques, techniques to generation
deformation invariant features, feature selection.
Machine Learning (Stanford Online Course)- Linear and Logistic Regression, Neural Networks, SVM,
Unsupervised Learning, Deep Learning, Anomaly Detection, Large-scale Machine Learning
Embedded Microcontroller Systems(33FJ128MC802)- PIC microcontroller and its applications in motor
control, PID controller etc
Digital Communication- Optimum receivers, carrier and symbol synchronization, Digital Communication
over band limited channels, ISI, channel equalization and channel coding.
Wireless Communication- Study of Rayleigh Fading and Ricean Fading Channel models, Multipath Effects,
OFDM, CDMA,TDMA, FDMA etc
Digital Signal Processing
Statistical Signal Processing
Contact this candidate