Adarsha Desai, San Bruno, CA, ********@***.*** https://www.linkedin.com/in/adarshadesai +1-213-***-**** Proficient software engineer with over 3+ years of experience, strongly motivated to build intelligent systems using deep learning and computer vision.
EDUCATION:
University of Southern California, Los Angeles Jan 2016 - Dec 2017 Master of Science in Electrical Engineering
Courses: Machine Learning (CSCI567), Image Processing(EE569), Probability(EE503), Multimedia System Design (CSCI576) BMS College of Engineering (Autonomous under VTU), Bangalore, Aug 2008 - June 2012 Bachelor of Engineering in Electronics and Communication Engineering TECHNICAL SKILLS:
Languages: C, C++, Matlab, ASCET, Python.
IDE and platforms: Visual Studio, Eclipse, Google cloud Platform, HPC, Code composer studio, windows, Linux. Tools: Git, CMake, ClearQuest, RTRT, SVN, DOORS.
Others: Keras, Theano, Torch, Tensor flow (deep learning frameworks), OpenCV, numpy, scipy, scikit-learn, pandas, VTK. PROFESSIONAL EXPERIENCE:
Computer Vision Intern, Intel Corporation, Santa Clara May 2017-Aug 2017 Worked on tracking and mapping part of SLAM algorithm. Created a data visualization tool for visualizing map data, which reduced the analysis time for the team by 4hrs/week (using VTK, C Senior Software Engineer Aug 2012-Nov 2015
Robert Bosch Engineering & Business Solutions Ltd, Bangalore, India.
• Developed Diagnostic (failsafe system) and Sensor Signal Processing software for Bosch’s safety system Electronic Stability Program
(ESP) in a Global team (India, Japan, USA) for Honda, Toyota & Nissan.
• Worked in Bosch Japan, Yokohama, for three months on challenging assignments from Honda. Roles and responsibilities (Programming in C, ASCET):
• Requirement analysis, Design and development of the Diagnostic software as per customer spec.
• Unit and module testing by simulation and System level testing using Hardware in Loop.
• Assistant Lead for the team of Diagnostic Responsible.
• Active member of Ideation team (Patents team), trained and mentored new college hires in the organization. ACADEMIC PROJECTS:
3D CNN for classification of 3D medical images (implementation in Keras, Python):
• Pre-processing of 3D SPECT images(79x95x64) of brain with spatial normalization using SPM8 tool box in Matlab, followed by Gaussian smoothing. Different shallow learning algorithms were tried for classification.
• Built a deep 3D CNN for Binary classification of the Pre-processed images, tried different architectures.
• Training on 1200 images, validation and testing on around 300 images was done (total data >2GB). Machine Learning algorithms (Implementation in MATLAB/Python):
• Implemented KNN and Naïve Bayes with cross validation on UCI ML Dataset test accuracy=72%
• Implemented Linear, Ridge Regression with cross validation on UCI ML Dataset .
• Experimented with Linear and Kernel SVMs using LIBSVM on UCI ML Dataset, test accuracy=95.55%.
• Implementation of K-means, Kernel-KMeans, GMM using Expectation-Maximization.
• Participated in Byte Cup 2016, problem solving using recommender systems. LeNet-5- Convolutional Neural Net (implementation in Torch):
• Image classification on MNIST dataset for positive and negative images with improved modified architecture.
• Classification for images with background from MNIST. Improved the network for translation invariance for MNIST. Real-time hand tracking, shape classification (implementation in C on TI Da-Vinci DSP board):
• Real time hand tracking using dynamic mean estimation, and shape detection drawn by user. Feature extraction using ORB descriptors. Bag of words representation obtained using k-means, classification using SVM. Image Processing (Implementation in C/C
• Implemented Non-Local means filter (NLM)for image denoising, histogram equalization, error diffusion and dithering.
• Image clustering based on texture analysis using Laws filters, PCA and K-means clustering.
• Image matching using SIFT and SURF, Image matching using Bag of Words of SIFT features (Open CV). Video summarization and image query (Implementation in C/C
• A synced AV player is made from. RGB and .WAV files. The highlights of this video were summarized in a shorter video. A given image is searched in the video and the closest matched frame is displayed. Interests:
I like hiking and have trekked through some challenging places such as the Himalayas and Mt. Fuji. I am also a fan of soccer and cricket.