Prateek Gaddigoudar

Software Engineer, Machine Learning and Computer Vision Enthusiast, C++, Python | Experience building ML/DL models and Automotive software in C

Biomedical Image Guidance Lab, CMU 0.6 Yrs , Robert Bosch Engineering and Business Solutions 1.11 Yrs , ECE Department, CMU 0.3 Yrs

Carnegie Mellon University, Pittsburgh Master of Science Electrical and Computer Engineering, B.V.B College of Engineering and Technology, Hubli Bachelor of Engineering Electronics and Communication

Software Engineer 02.00 Yrs, Machine Learning Engineer 00.06 Yrs, Software Development Engineer 0-1 Yrs,

C,  C++,  python,  Machine Learning,  Computer Visio


About Prateek Gaddigoudar

I developed a keen interest in Machine Learning and Computer Vision when I was researching “Pedestrian Detection and Tracking using Particle Filtering” during my undergraduate program. The idea that accumulated data of pedestrians could be used to make predictive models for their positions amazed me-- so much so that I ended up publishing a paper on it.

(https://ieeexplore.ieee.org/document/8229782)

Given my knack for Physics and Mathematics, I was driven to major in Electronics and Communication for my undergraduate degree. But even then, I was inclined towards real world problems.

I interned with Bosch and then I joined them full time as a Software Engineer for 2 years. I was primarily responsible for design, development, and testing of functional units in Powertrain ECU related to communication of diagnostic data.

During my stint with Bosch, I also got to put my coding skills (C, C++, Python) to use. But to get a thorough understanding of Machine Learning, Deep Learning and their application in autonomous vehicles, I decided to pursue my Master’s from Carnegie Mellon University, Pittsburgh.

I have been working as a Computer Vision and Deep Learning Intern at Biomedical Image Guidance(BIG) Lab situated in Robotics Institute, CMU under the mentorship of Dr. John Galeotti since April 2021. My work here includes carrying out tasks such as Segmentation, Detection and Tracking using several Deep Learning and Graphical Models on Ultrasound images.

The experience has been an enriching one. I have been fortunate to study and TA under the finest professors and work alongside some of the best minds. I aspire to continue making meaningful contributions in the domains of AI/ML or Software development in the long run with my skills and expertise.

Total Work Experience: 03.06 Yrs

Domain Experience : Automative domain,Biomedical Engineering Domain

Work Sponsorhip/Visa Required? : Yes, F1 OPT

PRIMARY AREAS OF EXPERTISE

  • C
  • C++
  • python
  • Machine Learning
  • Computer Vision

ADDITIONAL SKILLS

  • Agile Development
  • Java
  • AWS
  • MySQL

Experience

  • Software Engineer 02.00 Yrs
  • Machine Learning Engineer 00.06 Yrs
  • Software Development Engineer 0-1 Yrs

WORK HISTORY

Company Name: Biomedical Image Guidance Lab, CMU

Job Title: Computer Vision and Deep Learning Intern

Experience: 2021 Apr - 2021 Oct (0yr 6mos )

• Carrying out research project for AI understanding of ultrasound image and surgical robot control using Bayesian Deep Learning.


Company Name: Robert Bosch Engineering and Business Solutions

Job Title: Associate Software Engineer

Experience: 2017 Jul - 2019 Jun (1yr 11mos )

• Developed software for Engine Control Module (ECM) under Powertrain Systems for effective communication of data related to vehicle diagnostics between ECM and Tester device.

• Awarded best performing fresher out of ~1000 employees for excellent display of coding skills and knowledge on Client-Server architecture, OSI model and communication channels such as CAN, Flexray, Ethernet and deep understanding of AUTOSAR in 2017.


Company Name: ECE Department, CMU

Job Title: Teaching Assistant

Experience: 2021 Jan - 2021 Apr (0yr 3mos )

• Assisted professor in running course “Principles and Engineering Applications of AI” by creating assignments, conducting recitations and bootcamps for students on Python coding and several AI concepts such as Bayesian Networks, HMMs, NLP, Robotics and Intelligent Agents.


Academic Projects

Title : Comparison and Synthesis of Musical Audio Source Separation Methods
Technology Used : Python, TensorFlow, Dimensionality Reduction, LSTM, CNN, Libraries - Pandas, Librosa, NumPy

• Employed various machine learning based techniques such as PCA, ICA, NMF and deep learning techniques such as Bi-directional LSTM, Wave-U-Net to separate musical audio into its source components.

• Measured several objective metrics to compare performance of these techniques on DSD-100 dataset. Through analysis, empirically showed that Wave-U-Net model outperformed other techniques with an improvement in signal-to-distortion ratio by +0.98 dB.


Title : Semi-Supervised Deep Geometric learning for COVID-19 Regional Severity Classification
Technology Used : Python, Graph Convolution Networks, Libraries - NetworkX, NumPy, Matplotlib

• Developed data-efficient Graph Convolution Network (GCN) to classify the COVID cases in 48 contiguous states of USA according to 5 severity categories, which achieved 73% accuracy using only 50% of labelled training data.


Title : Musical User Preference Modelling
Technology Used : Python, Unsupervised Learning, Weighted K-Means, GMM, Libraries - Spotipy, Librosa, NumPy, SciPy

• Implemented a novel music recommender based on acoustical features of song unlike traditional methods of suggestion through genre, enabling users to listen to songs across the globe which sound similar to their music palette using unsupervised machine learning.

• Procured user listening history from Last.fm dataset containing 10K songs, and extracted acoustical features of songs from Spotify API to train Weighted K-Means Clustering model with a precision of 0.606 (beating other unsupervised techniques).


Title : Multi-Face Detector
Technology Used : PyTorch, Adaptive Boosting, Non-Maximum Suppression

• Implemented AdaBoost based multi-face detector trained using Eigenfaces computed on Yale Face Database with 93% accuracy.

• Employed Non-Maximum Suppression (NMS) post-processing algorithm to merge bounding boxes belonging to the same faces.


Title : Pedestrian Detection and Tracking using Particle Filtering
Technology Used : Python, OpenCV, PIL, MATLAB, HMMs

• Researched different modelling and prediction techniques, including Kalman filter and Particle filter. Realized the algorithms using OpenCV library to perform Pedestrian Tracking in complex environments involving moving backgrounds and mutual occlusions between pedestrians.

• Authored paper “Pedestrian detection and tracking using Particle filtering” published in ICCCA 2017, which demonstrated that Particle filter performed better than Extended Kalman filter under challenging non-linear conditions on real-time videos acquired from Caltech database.


Education


, Electrical and Computer Engineering



, Electronics and Communication