Jose Joy

Jose Joy

SDE 2, Amazon AI

Contact Me

About Me

Currently working as an individual contributor in the Amazon SageMaker team. I have a Master's Degree from UC San Diego, specializing in Machine Learning and Data Science.

My professional interests include software development, distributed systems and machine learning.

My hobbies include tutoring, reading, movies and cooking.

Work Experience

Software Development Engineer 2 - Amazon Sagemaker (June 2019 - Present)

SageMaker Processing.

Graduate Teaching Assistant - ECE Department, UC San Diego (Sep 2018 - Mar 2019)

Teaching assistant for Statistical Learning (ECE-271 A)

Teaching assistant for Pattern Recognition and Machine Learning(ECE-175 A)

Graduate student researcher - Statistical Visual Computing Lab, UCSD (Feb 2018 - August 2018)

Built a machine learning model to estimate relative size of objects from an RGB image. The model uses a mask rcnn backbone to identify and localize objects, and to extract feature. A siamese network was used to compare the relative size of objects. Since there were no datasets available for this task, created an entire dataset pipeline to scrape images, processsing and to annotate the data using amazon mechanical turk

Reseach Intern - HTIC, IIT Madras, India (Sep 2016 - March 2017)

Designed a machine learning algorithm for motion sensing in the field of tennis analytics and performance enhancement. Using our dataset, classified tennis shot types and predicted ball speed generated and sweet spot accuracy. The algorithm can give targeted suggestions to players based on data collected from a 9 axis motion sensor, to improve their play.

Designed a machine learning algorithm for detecting stress, using Electrodermal activity (EDA) and Heart rate variability(HRV). The algorithm can be used to detect the onset of stress, and classify psychological and physical stress. With certain improvements, the algorithm can be extended to predict seizures and alert the user.

Research Intern - Intel Labs, Bangalore, India (Jan 2016 – June 2016)

Developed a smart sensing algorithm using Compressive Sensing to improve network capabilities of Wireless Sensor Networks and IoT. Using only a subset of data from a sensor field, the data corresponding to the entire sensor network was reconstructed with good accuracy. Data acquisition from a subset of sensors would lead to lower power consumption and increased lifetime. The algorithm was also able to detect onset of anomalies and change operation flow accordingly.

Collaborated on an Invention Disclosure Form titled ‘Novel Compressive Sensing Scheme for Power Efficient Data Aggregation in a Spatial IoT Wireless Sensor Network.'

Summer Intern - Systemantics India Pvt. Ltd., Bangalore, India (June 2015 - July 2015)

Designed an adaptive motion sensing algorithm to determine the configuration of a robotic arm using MPU 9150 sensor. Instead of using an expensive absolute encoder for determining configuration of a degree of freedom, was able to use a cheaper incremental encoder and motion sensor to accurately find the angular displacement from a fixed point, which would lead to lower build costs.

Summer Intern - C-DIT, Trivandrum, India (May 2014 - July 2014)

Designed and developed websites for various functional bodies of Government of Kerala, India, using Joomla content management system

Projects

Image captioning and retrieval

Created a system that can generate captions from an image and can recommend similar images from a dataset or Google images. The system uses an image captioning network based on Resnet 152 encoder and a single layer LSTM based decoder. This system can be used to create an entire dataset of images, from few seed images.

Handwritten mathematical expression evaluation

Used various machine learning algorithms like Convolutional Neural Networks, Multi Layer Perceptrons, AbaBoost and Random Forest to implement a system to evaluate handwritten mathematical expressions. The system works by locating and identifying digits and math symbols from an image of handwritten mathematical expressions, and processing them to evaluate the expression.

IMDB Data analysis and visualization

Designed a system to analyze and visualize the IMDB dataset that contain information about 4.3 million titles and 8 million artists. The system can be used to obtain hard to collect information like popularity of a particular genre or a director throughout the year, most popular movies of a particular year, average rating and popularity of a particular genre etc.Also designed a movie recommender system based on collaborative filtering.

Fake news classification

Designed a classification model that checks whether is news headline is fake or not. The model uses n-grams, tf-idf, non-negative matrix factorization and Latet Dirichlet Allocation to check if a news headline is similar to that of fake news.

Recommender system for local businesses

Designed a recommender system for suggesting businesses to customers. For this task, used google local business data which had over 200,000 reviews from 20,000 users, for 19,000 businesses. Used various techniques like matrix factorization, collaborative filtering and latent factor models to predict if a user would visit a business and the likely rating the user would give to that business.

Word suggestion

Designed a word suggestion program using n-gram mixture models. The models tries to suggest next likely word, given a sequence of words using a weighted combination of a unigram, a bigram and trigram model and was trained using text data from various open sources.

Portfolio Selection algorithm

Implemented a portfolio selection algorithm, using Cover's universal portfolio theory. The algorithm works by trying to track the best possible constant rebalance portfolio, on a set of stocks. To run the algorithm, stocks from 14 companies were downloaded from Yahoo! finance.

Rock Paper Scissors agent

Developed an agent to play rock paper scissors, using the concept of universal probability. The agent tries to estimate the underlying probability distribution of the opponent to defeat the most probable move.

Other projects

Lane detection

Designed an algorithm to detect road lanes in an image. The algorithm is capable of suppresing noise, enhances image contrast, uses canny edge detection to extract edges and Hough transform to identify lane separation lines.

Image segmentation

Designed an image segmentation algorithm using a combination of k-Means clustering, Gaussian Mixture Models and Bayesian decision theory. The algorithm works by fitting gaussian mixture models of objects and background, using Expectation Maximization algorithm, and using the mixture models on a Bayesian classifier.

Implementation of various machine algorithms using numpy

Implemented various machine learning algorithms like multi-layer perceptron, adaboost etc. on numpy to classify digits from the MNIST dataset.

Fuzzy logic based control system

Developed Fuzzy Logic based Smart Temperature Control System using Intel 8086 processor and performed simulation of the assembly using Proteus software.

VLSI design

Designed differential telescopic operational amplifiers and digital logic circuits using cadence virtuoso (180 nm technology).

Activities and Achievements

• Perfect score in GRE Quant, July 2016

• Ranked within the top 1 percentile in Electronics and Instrumentation National Level Graduate Aptitude Test (GATE), February 2016

• Ranked within the top 1 percentile in IIT Admissions Exam and All India Engineering Entrance Exam, May 2012

Courses

University of California San Diego

• Statistical Learning I (Generative modelling; grade B+)

• Statistical Learning II (Discriminat methods; A)

• Statistical Learning III (Deep Learning)

• Recommender systems (A)

• Linear Algebra (A)

• Image Processing (A)

• Programming for Data Science (A)

• Probability & statistics for data science (A-)

• Convex Optimization