Sneha Sudhir Shetiya

Studying at North Carolina State University
Living in United States

About

A keen learner and a passionate individual with high expertise and experience in the Electrical Department of Mercedes Benz Research and Development India. With the spirit of learning and to dive deep into Machine learning and Computer Vision domain, I am currently pursuing my Masters in North Carolina State University.
Following is my skillset:

-Knowledge of Automotive /Powertrain Domain
-Diagnostics(ISO 26262 standards) and CAN(ISO-11898)
-VBA/Python/Matlab tool development
-Matlab Simulations and Analysis
-Static Verfiication Analysis using Polyspace
-Verification and Validation of Software
-Efficient in Vector tools CANdelaStudio,CANDiva,CANoe
-Configuration management and Requirement Gathering in MKS Integrity,Jama
-Process flow for ISO26262 functional Safety
-Worked with Agile methodology-Jira
-Expertise in Jenkins tool
-Complete V-cycle software development Expertise.
-Neural Networks(CNN,RNN,SVM,GAN techniques)
-Computer Vision algorithms
-Perception, Path Planning and Motion Control Topics for ADAS Vehicles
-Sensor Fusion LIDAR/Camera/Radar

Experience
NVIDIA
Automotive Reqirements/System Engineering Intern
Company NameNVIDIA Internship
Dates EmployedMay 2020 – Aug 2020
Employment Duration4 mos
LocationUnited States
Orchestrated decomposition for 35 Tier 3 Projects of DRIVE OS at Nvidia Drive on the Jama Platform into Stakeholder Requests, Architectural Decomposition and Features. Generation of excel Reports using Birt and Velocity APIs.
Involved in REST API scripts to develop tools and scheduling of these for automation within Jama.
Communicated and cross-coordinated with various teams in the Bay Area, India and Germany to make decisions pertaining to process management in Autonomous Driver Assistance Projects.
Drive topics of Automotive Glossary within all Projects.Carved out detailed plan including Impact analysis in order to achieve the Project.
Duplicated all Projects utilizing Branching across 3 branches of software releases for efficient handling of customer requirements.
Mercedes-Benz Research and Development India
Total Duration5 yrs 2 mos
TitleSenior Product Design Engineer
Dates EmployedApr 2019 – Aug 2019
Employment Duration5 mos
LocationBengaluru Area, India
-Global visibility to NAFTA and European colleagues in co-ordination and configuration management activities for Engine and Aftertreatment ECUs.

-Verification and validation using static code analysis of C code with Polyspace tool for frames of Engine and Aftertreatment ECUs.

-Experience in m-script,batch script and vba for automation activities.

-Courses in Machine Learning,Neural networks , deep learning, R Programming and SAS Programming done during this period due to interest in AI topics.
…
see more
TitleEngineer
Dates EmployedApr 2015 – Mar 2019
Employment Duration4 yrs
Development and maintenance of tools used for the documentation and certification of OBD protocols for European, NAFTA and Tier 4 variants for HDEP and MDEG , ACM and MCM ECUs
TitleGraduate Engineer Trainee
Dates EmployedJul 2014 – May 2015
Employment Duration11 mos
Testing and validation of services for ISO 14229 protocol for ACM and MCM ECUs using manual based approach and automated Diva testing based approach.
Projects
Cyclist Detection using Detectron2 model
Using the KITTI Research Suite's cyclist dataset, our team implemented Facebook AI's Detectron2 model to detect cyclists in still frames taken from moving vehicles. Occlusion and distance were the main challenges of this Project and those results were explored.
Soybean detector using CNN and Resnet Approaches
Classification based on leaf wilting in soybean plants using neural networks. Used the architecture of Convolutional Neural Networks along with VGG. The dataset proved to be unbalanced. Thus tried to improve accuracy by using the architecture of resnet and varying the weights.Applied data augmentation. The images provided as inputs had plants with different stress levels that were classified as per the amount of change in the wilting.
Extraction of Harr features and Implement Adaboost algorithm
Extract Harr features for the FDDB-folds dataset. Calculate classification error for each weak learner and draw the ten best features. Decision tree of depth 1 has been used to do the same.
The prediction of Adaboost is done by computing the weighted average of the predictions of all weak learners.
Face Image Classification
Modeled Gaussian, Mixture of Gaussian, t distribution,mixture of t-distribution, Factor Analysis and mixture of Factor Analyzer involving data preprocessing and cross validation strategy.
Utilized Expectation Maximization algorithm using hidden variables for improved modelling.
1000 face and non-face images for training and 100 test images of size 10*10 from FDDB dataset were used on the various models.
Train an MLP for digit recognition using MNIST dataset
Implemented back propagation algorithm to Multi Layer Perceptron for handwritten digit recognition with a test accuracy of 90.61% with tuning hyper-parameters like number of layers and learning rate on 3000 samples in training, 10,000 samples in validation and 10,000 test samples.
Laplacian Blob Detector

Used Laplacian of Gaussian filter and maximum suppression techniques to create blobs on an image image based on the internsity parameters.
Design a convolution function of an image using six different types of filters of Box,Prewitt,Sobel and Roberts.Implement padding for the various types of clip/zero padding,wrap around,reflect across edge and copy edge.
Visualising and checking the smoothness of images by finding the intensity differences implemented for Gray,Lab,HSV ,Hue and color images.Using djshiktra algorithm and network graph concepts, finding the shortest path between two pixels in n image.
Image smoothing and shortest path detection