Biography

Amit Maraj is the program co-ordinator of the Artificial Intelligence, Analysis, Design and Implementation graduate certificate and a full-time professor within the Computer Systems and Programming streams at Durham College (DC) in the School of Business, IT & Management.

Maraj is an artificial intelligence (AI) and machine learning enthusiast with significant expertise in natural language processing (NLP) and deep learning. His areas of research are advancing and Medium Enterprises (SMEs) by providing unique and cutting edge technological solutions using AI. He has worked with sectors including agribusiness, health care, transportation, and health and safety.

As a principal investigator with DC’s Office of Research Services, Innovation and Entrepreneurship (ORSIE), Maraj has developed and implemented more than 15 industry-led applied research projects focused on helping organizations adopt AI technologies into their business models.

With ORSIE, he helped establish the Hub for Applied Research in Artificial Intelligence for Business Solutions (AI Hub), and created boot camps, workshops and several fully documented machine learning examples, which are hosted publicly for use by students.

A member of the Program Advisory Committee for the Data Analytics graduate certificate and the Artificial Intelligence Analysis, Design and Implementation graduate certificate, Maraj has been awarded with the Silver Faculty of Excellence in Leadership by Colleges and Institutes Canada as well as the Outstanding Faculty Researcher Award by DC.

In addition to DC, Maraj works as chief executive officer and founder in Oiika, and as a teaching assistant for Ontario Tech University for Machine Learning, Cryptography and Network Security, and Discrete Mathematics courses.

Maraj is completing his Doctor of Philosophy in Computer Science with a specialization in NLP and holds a Master of Science in Computer Science and a Bachelor of Science in Information Technology Security (Honours), from Ontario Tech and a Computer Systems Technician Diploma from DC.

Additional roles at Durham College

  • Program Coordinator: Artificial Intelligence, Analysis, Design and Implementation Graduate Certificate
  • Member, Program Advisory Committee, Data Analytics Graduate Certificate (2017-18)
  • Member, Program Advisory Committee, Artificial Intelligence, Analysis, Design and Implementation Graduate Certificate (2018)

Area of Research

  • Advancing Agriculture, Agribusiness, and Tourism
  • Enabling Technologies of the Future
  • Cultivating Healthy Lives and Resilient Communities
  • Enhancing Scholarly Teaching and Learning

Research Interests

  • Machine Learning
  • Data Science
  • Artificial intelligence
  • Deep Neural Networks
  • Natural Language Processing
  • Recommender Systems
  • Web Development
  • Mobile Development
  • Language Modeling
  • Natural Language Understanding
  • Dev Ops

Credentials/ Education

  • Doctor of Philosophy in Computer Science, Ontario Tech University, expected graduation 2022
  • Master of Science in Computer Science, Ontario Tech University, 2018
  • Bachelor of Science in Information Technology Security (Honours), Ontario Tech University, 2016
  • Diploma in Computer Systems Technician, Durham College, 2014

Current Research projects

Optical Character Recognition (OCR) Pipeline for Receipts

Industry Partner: The Mobile Experience Co

A solution was required to improve the time needed to sort through the industry partner’s receipt submissions from users for promotions the company ran. The team at DC including four student research assistants, led by Principal Investigator Amit Maraj, is developing a four-stage pipeline to pre-process, perform OCR, extract text, and draw insights from pictures of receipts fed into the MobileXCo system.

Customer Service Bot

Industry Partner: Precise Park Link

As the largest parking management and equipment provider company in Canada, this industry partner deals with several hundred parking lots across the nation. They needed a solution to the large number of customer service requests at the parking lot gate. Co-Investigators Amit Maraj and Clint MacDonald along with seven student research assistants developed a Customer Service bot, which can replace up to 70% of these requests. The pipeline for the bot includes speech-to-text processing, a recurrent neural network, and a text-to-speech synthesizer for an end-to-end solution.

eCommerce sizing and recommendation using AI

Industry Partner: TelliGram Company

TelliGram Company is an Augmented Reality (AR) company focused on providing low cost, affordable Augmented Reality experiences for businesses and consumers. With Durham College they wish to develop a personal assistant for their online shoppers using AI who would provide consumers with product visualization (AR), actual sizing (AI) and recommended shoes (also using AI). Students under Amit Maraj’s and Noopa Jagadeesh’s supervision are developing and training neural networks so that they can use harvested feet and shoe images to create a Computer Vision pipeline in order to accurately detect foot and shoe dimensions. The proposed pipeline will then use those dimensions to create shoe and shoe size recommendations.

Past research projects

Remote monitoring health app – Health Espresso

Industry Partner: iCare Home Health Services Inc.

Artificial intelligence (AI) was built into an app to provide medication management, real time tracking of body readings and virtual caregiver reminders. Under the direction of the Principal Investigator, Amit Maraj, four student research assistants (two went on to become Principal Investigators) were involved in learning the user’s behaviours to enhance the application’s experience through the development of algorithms on speech recognition and natural language processing. Several unique AI- related features were added to the Health Espresso app to enhance accessibility and ease of use. A medical care specific speech recognition and classification algorithm allows users to command the Health Espresso application with their voices instead of through the touchscreen interface.

Speech-enabled Assistive Mobile Application
Industry Partner: Sultan of Samosas Historically recording metrics on paper, the industry partner worked with the Co-Investigators Amit Maraj and Clint MacDonald to develop a mobile application to integrate into their current daily operations. The four students involved in the project researched and developed a natural language processing assistant accessible throughout the mobile application which can respond to, and perform user-specified tasks. The assistant was created with several layers of natural language understanding to account for noisy environments and varying accents.
Speech Recognition for Closed Captioning using Artificial Intelligence

Durham College identified a need for efficient and customized closed captioning of teaching materials including video lectures. As the Principal Investigator, Amit Maraj scoped out a project to use AI to solve this issue. Under Amit’s supervision, students developed a software to run on various Durham College (DC) teaching materials for accurate closed captioning. This software was wrapped in a web application and featured robust editing and per-word highlighting based on confidence.

AI-integrated Mobile Application
Industry Partner: We Traq Inc. Located in Pickering, WeTraq Inc. engaged with the AI Hub at Durham College (DC) to integrate AI into their tracking application to better monitor individuals at risk of wandering. The research team, including two student research assistants, focussed on a specific demographic that requires constant attention as a result of their mental health challenges, such as memory loss or diminished mental acuity. Smart algorithms were developed to analyze historical location patterns of WeTraq device users, which could then send alerts and warnings to caregivers in the event of unusual activity. The algorithm developed incorporated a variety of complex, dynamic Recurrent Neural Networks – an artificial network often used in Natural Language Processing applications.
Microbiome profile management Using AI
Industry Partner: uBioDiscovery This project created a scalable and consistent database to store all uBioDiscovery’s client treatment plans and information. This data was then used to train and advance the AI solution. The PI and two student research assistants developed AI pipeline consisted of several data science, machine learning and deep learning paradigms including Data Cleaning, Data Mangling, Data Augmentation, Data Preprocessing, Random Forests, Decision Trees, and Deep Neural Networks. The core AI models were then trained on several hundreds of data points to produce an accurate probabilistic distribution to help augment the final predictions made by the in-house Biologists. A web portal was also built, which captures clients’ health status after their diagnosis and automatically generates a treatment plan (e.g., an appropriate diet based on the gut-bacteria results).
Image Recognition for Automatic Image Caption Generation

Using state-of-the-art Artificial Intelligence (AI), this research project involved the wide space of Image Recognition. Under the direction of the PI, students researched artificial neural networks such as multilayer perceptrons and convolutional neural network architectures to develop an AI-based solution for generating descriptive captions for DC social media posts based on what the AI learned from the picture in question.

Unmanned Aerial Vehicle (UAV) for Precision Agriculture

Industry Partner: Woodleigh Farms Ltd.

Woodleigh Farms Ltd., uses UAVs to inspect crops with normalized difference vegetative index (NDVI) and near-infrared (NIR) sensors. Their challenge was the inability of the software to incorporate and process the NDVI images into the farm software programs. This project, with the assistance to two student research assistants, developed the bridging software (middleware) to process the data from the UAV and incorporate the high resolution NDVI images into a variety of databases allowing Woodleigh Farms Ltd. to scout fields using different methods with a high level of accuracy.

A Buyer and Supplier Platform (Food Portal)
Industry Partner: BBH Ventures The industry partner, BBH Ventures Ltd. required a solution to address the waste of upwards of 40 per cent of all food produced. An online bid-based platform was developed to host a network of food producers and suppliers to optimize the sale and delivery of surplus time-sensitive food products. By connecting time-sensitive products to buyers in real time, Shelflifefoods.com (SLF) has established a niche market that aims to capture significant capital losses from food waste.
Mobile Application Mirrored Web Platform
Industry Partner: Think Dirty As a result of growing public concern over the safety of cosmetics products, the company, Think Dirty required a mobile app to address a large gap in the provision of toxicity, carcinogenicity and allergenicity information to consumers. A Mobile Application Mirrored Web Platform was developed that allows consumers to acquire the information they need to make informed choices and better manage their health. Think Dirty received direct economic benefits through the creation of novel intellectual property.
Pitstop-Dashboard Management System
Industry Partner: Ansik Inc. Ansik Inc. developed a service platform, Pitstop, powered by aggregating data from multiple sources to perform machine learning and data analytics that result in real-time diagnostics and predictive maintenance. A functional dashboard with dealership metrics and statistics was developed to provide the company with more control over interactions with their customers and other stakeholders.
Backend Task Automation
Industry Partner: Deveron UAS Ltd. Deveron UAS is a drone data service solution provider for agriculture that helps farmers to increase farming efficiencies and reduce farming costs. By integrating a cloud storage solution into their daily operations along with a pre-existing web-based application that interfaced with all their image data, Deveron UAS was able to significantly increase their operational efficiency and service delivery goals.
A software for greenhouses
Industry Partner: Floragenie Accurate scheduling and detailed record keeping is crucial for small and medium sized greenhouse growers as various factors impact their ability to be able to grow plants to marketable size at the right time of year. A web based software program was developed to simplify and expedite planning and productions of mid-tier greenhouses.
High School Student Alcohol Consumption Study

In this project a publicly available dataset of over 500 students’ alcohol consumption habits was analyzed, sanitized, cleaned, and pre-processed. Subsequently, several unsupervised Machine Learning algorithms were employed to yield correlations within the dataset that revealed students who consume more alcohol have a positive correlation with more free time on their hands, spend less time studying, and lack guardians whose occupations reside within health care.

Full Stack Web Development

Simultaneously completed multiple projects for Konkussion Inc. documenting all research findings within a given time frame. This included writing efficient and optimized code in Node.js and NoSQL for multiple backend services, such as real time PDF generation and MongoDB collection modelling and interaction. Front end tasks included taking care of aesthetic design implementations and tweaks and overhauls on company’s main webpage, all performed with Pug (Jade), CSS, Bootstrap and JavaScript.

Technologies and Frameworks used

  • Python
  • Jupyter Notebooks
  • TensorFlow
  • Keras
  • AWS
  • JavaScript

Researcher Presentations/ Publications

Sept. 2018 How Artificial Intelligence Can Improve Your Competitive Advantage, 2018 Innovation Forum, Ajax-Pickering Board of Trade (APBOT)

June 2018 Introduction to Artificial Intelligence, DC Professional Development Day &

Introduction to Artificial Intelligence, DC Faculty Retreat

May 2018 Applied Artificial Intelligence in Business, DC Board of Governors

March 2018 Using Seq2Seq Recurrent Neural Networks for Deep Learning Chatbots, Ontario Tech University

Sept. 2017 AI Solutions in Health Care, Indian Institute of Technology Alumni Canada (IITAC) Lecture Series

June 2017 Deep Learning, Ontario Tech University 4th year IT students

  • Created several fully documented Neural Network examples which is hosted publicly on GitHub for use by students in the

March 2017 Machine Learning session at 36 hour Hackathon, Ryerson University

  • Created several fully documented Machine Learning examples, which is hosted GitHub for use by students in the

Awards/ Recognition

  • April 2018 - Faculty of Excellence in Leadership (Silver), Colleges and Institutions Canada
  • May 2017 - Top Faculty Research Award, DC, Research Day