(For Githubbers !) Forked knowledge from all the those who knows anything better than me. Trained sequential model named Darshil Trivedi with all kind of Data Augmentation to get a better accuracy. Note that overfitting was always kept in mind and was avoided.
(To rest of the world !) Darshil Trivedi is currently working as Field Applications Engineer specialist at ADI. His focus is towards Embedded Systems and is working as a bridge to connect Hardware with Software. He Graduated from University at Buffalo, completing his M.S in Electrical Engineering. He did his Undergraduate in Electronics Engineer from The Maharaja Sayajirao University of Baroda-India. His primary interest lies in Embedded sensor technology and Analog/Digital designing. He designed and coined an output capacitor-less low dropout voltage regulator in 65-nm CMOS Technology. Getting influx to the world of Machine Learning he is implementing different algorithms from scratch and employed Convolutional Neural Networks for Image Recognition. As he is having a passion for FPGAs, he is also into VHDL and Verilog coding. During his undergraduate, he worked with ADC-DAC, Pulse-width Modulation, Analog/Digital sensors (Light Dependant Resistor, Thermocouple, IR etc.) and PD controller built from Op-amps. He is acquainted with software like Cadence, Matlab, Vivado, LabVIEW, Eagle PCB designing.
As a Field Applications Engineer (FAE), working as a bridge by connecting end customers to Analog Devices technology.
Providing help in system level designing to customer projects related to embedded systems and MEMS technology.
Working with customers throughout the life cycle of their product/project and help them in creating next gen technology.
Test and verify evaluation boards/systems to create a better user experience for customers..
Working with application engineering team, evaluated & tested specification on datasheets and software user guides.
Developed python API, MATLAB toolbox and example codes of ADI products for internal product line team and System Development Group.
Collaborated with other new college graduate to perform a full embedded product development which included designing a smart chair to determine the ergonomics of a person.
Provided technical support to customers with the chip recommendation, optimum use of the chip and design review.
Worked with other FAEs on Audio Automotive Bus(A2B) to improve in-cabin audio and user experience.
Built a FPGA Calculator using Basys3 board. VHDL as Language in Vivado tool was used for the project. The Calculator was able to perform simple Arithmetic operations as well as logic left shift between 2 decimal operands entered using slider switches. The operands and results were displayed on 7-segment display. Different push buttons on basys3 were used to store operands and select operations to be performed. Note the main code is password portected and password for it is 1234.
Implemented Perceptron, SVM, Linear & Logistic Regression, k-NN, Random Forest and K-means from scratch on MNIST and Fashion MNIST datasets and applied 10-fold cross validation to get a maximum accuracy. Employed a CNN based model for Image Recognition using the AlexNet architecture. Many tweaks were made in the existing CNN to get a better accuracy. Top-5 accuracy of around 68% was obtained for oxflower17 dataset. Reinforcement Learning, Q & double-Q learning, Transfer Learning and Convolution Neural Networks were the Machine Learning models being employed.
Tapeout of this work is expected in by september 2020. So stay Tuned!
Researched and designed an output capacitor-less low dropout voltage regulator in 65-nm CMOS Technology. Designed a Reference Voltage for LDO which was a Subthreshold Voltage Reference with scalable output voltage. Developed RTL-level logic and MOSFET-level layout to perform different analysis using Cadence Virtuoso. Significant improvement in Load & Line regulation, Temperature Compensation and lower Chip area was obtained compared to previous result, resulting in LDO for SoC usage.
Acknowledging the world of wearable and implantable Sensor tried to understand the sensors insmart-phone. Designed an Android Application to collect real time sensor data and test & verify the data.
Projects at Maharaja Sayajirao University of Baroda
For this we built LASER galvanometer scanner and closed loop Servo amplifier which reflects the laser beam to form continuous image on smoke screen. The Servo amplifier Board (PCB) was built using op-amps and was coupled to PD Controller circuit with capacitive feedback. Laser Intensity control circuit was built to control the depth of the 2D image on smoke screen.
Constructed a car using Bluetooth module HC-05 and Arduino which can be operated with smart phone or laptop. Interfaced 2 stepper motors in X and Y direction using 8051 microcontroller to draw different geometric shape on the graph lying on the chassis.
Liked my Work? (as Steve Jobs always said !) Design is not just what it looks like and feels like. Design is how it works. and my webpage works. So even you can fork it.