Vivekananda Ponnaiyan

Birla Institute of Technology and Science, Pilani

MSc, Physics

N/A – Present

Birla Institute of Technology and Science, Pilani

BE, Electrical and Electronics Engineering

N/A – Present

Stealth

NewCo.

January 1, 2022 – Present

Grabango

Director of Machine Learning and Computer Vision

January 1, 2020 – January 1, 2021

Chime

Head of Machine Learning & Data Science

January 1, 2018 – January 1, 2020

San Francisco Bay Area

BMW of North America, LLC

Team Lead - Deep Reinforcement Learning & Perception

January 1, 2017 – January 1, 2018

Mountain View, CA

Wave-Ai

A Independent Machine Learning Consultant

January 1, 2016 – January 1, 2017

San Francisco Bay Area

Medidata Solutions

Engineering Manager - Rave Safety Gateway

January 1, 2016 – January 1, 2017

San Francisco Bay Area

RadiumOne

Engineering Manager - Reporting and Analytics, DMP

January 1, 2015 – January 1, 2016

San Francisco Bay Area

FitMountain.com

Entreprenuer - 2 Companies - (CoachAce & FitMountain)

January 1, 2012 – January 1, 2015

San Francisco Bay Area

Bloomberg LP

Manager - Software Development

January 1, 2004 – January 1, 2011

Greater New York City Area

Brookhaven National Laboratory

Software Engineer

January 1, 2002 – January 1, 2004

Upton, NY

Nokia

Software Engineer

January 1, 2000 – January 1, 2002

Long Island, NY

Vehicle Detection & Tracking System (Computer Vision using sliding window technique)

May 1, 2017 – Present

The goal of this project was to create a vehicle detection pipeline. The steps that were taken: 1. Performed a Histogram of Oriented Gradients (HOG) feature extraction on a labeled training set of images 2. Train a classifier Linear Support Vector Machine vehicle classifier using the massive GTA data 3. Apply a color transform and append binned color features, as well as histograms of color, to HOG feature vector to make it more robust 4. Implemented a sliding-window technique and used the trained classifier to search for vehicles in images. 5. Create a heat map of recurring detections frame by frame to reject outliers and follow detected vehicles. 6. Estimate a bounding box for vehicles detected in each frame of a video to be able to track the same.

Customer Segmentation of wholesale distributor's clients using Unsupervised Learning (Clustering)

May 1, 2017 – Present

Used K-Means clustering to segment customers of a particular wholeshale distributor in order to better understand customer behaviors and to devise corresponding marketing strategy and operational plan for introducing new services to the market. The project included, data pre-processing, Principal Component Analysis (PCA), independent Component Analysis (ICA), Gaussian Mixture Model clustering (GMM) and K-Means clustering techniques.

Image classification using Deep Learning and CNNs (Convolutional Neural Networks)

May 1, 2017 – Present

In this project I build a Image classifier based on the CIFAR-10 dataset. We: 1. pre process the images to normalize them, 2. one hot encode the labels 3. create a tensorflow convolutional neural network graph 4. split data into a training set and a test set 5. create a model using the training set 6. improve the model by tweaking hyper-parameters, variable initialization values, addition/deletion of convolutional layers and fully connected layers. We then evaluate the training accuracy versus the test accuracy to figure out the number of epochs to run. Finally we evaluate the algorithm on photos found on the internet.

A Traffic Sign Recognition Classifier for Self Driving Cars

April 1, 2017 – Present

Self Driving Cars will have to recognize traffic signs and then obey the same. In this project we build a traffic sign classifier using a Deep Learning. Steps include: 1. normalizing image data 2. balancing the data set 3. building and training a convolutional neural network with a traffic sign dataset 4. tweaking the initializations, hyper parameters, fully connected layer sizes, the CNN filter sizes to increase the accuracy of the system 5. testing the accuracy of the system with images found on the internet

Predicting Boston Housing Prices with Linear Regression

April 1, 2017 – Present

Using historical price data of Boston house prices to create a predictive model. Project included data pre-processing, getting rid of outliers and evaluation of the predictive model.

Find Lane Lines after un-distorting images using camera calibration, perspective transforms and using color and gradient thresholding.

March 1, 2017 – Present

In this project we use many different Computer Vision techniques to come up with a system that can detect lane lines for use in an Autonomous Vehicle : The goals / steps of this project are the following: * Compute the camera calibration matrix and distortion coefficients given a set of chessboard images. * Apply a distortion correction to raw images. * Use color transforms, gradients, etc., to create a thresholded binary image. * Apply a perspective transform to rectify binary image ("birds-eye view"). * Detect lane pixels and fit to find the lane boundary. * Determine the curvature of the lane and vehicle position with respect to center. * Warp the detected lane boundaries back onto the original image. * Output visual display of the lane boundaries and numerical estimation of lane curvature and vehicle position.

Finding donors for a non-profit using Support Vector Machines & Random Forests.

March 1, 2017 – Present

In this project I employed several supervised algorithms to accurately model individuals' income using data collected from the 1994 U.S. Census. I then choose the best candidate algorithm from preliminary results and further optimize this algorithm to best model the data. The goal with this implementation was to construct a model that accurately predicts whether an individual makes more than $50,000 in a bid to create a target market for a non-profit. The dataset for this project originates from the UCI Machine Learning Repository.

Train a Smart Cab to Drive with Reinforcement Learning

March 1, 2017 – Present

In the not-so-distant future, taxicab companies across the United States no longer employ human drivers to operate their fleet of vehicles. Instead, the taxicabs are operated by self-driving agents, known as *smartcabs*. I implemented a Q-Learning algorithm for the self-driving agent to guide the agent towards its destination within the allotted time. Improved upon the Q-Learning algorithm to find the best configuration of learning and exploration factors to ensure the self-driving agent is reaching its destinations with consistently positive results.

Use Deep Learning to Clone Driving Behavior

February 1, 2017 – Present

In this project, clone driving behavior using deep neural networks and convolutional neural networks. One way of teaching a Neural network to drive a car is to train it with data collected by watching a human driving a car. In this projects I: 1. drove a car around a simulated race track, 2. recorded images and steering angles from driving around the simulated track, 3. built and trained a convolutional neural network using Keras & tensorflow using the above data, 4. regressed the steering angle using the images as input, 5. pre-processed data, augmented data and tuned the number of layers, size of filters of the neural network to improve performance of the system 6. tested the system by seeing how well it drives around the track on its own.