Artificial Intelligence & Data Science

The field of Artificial Intelligence (AI) and Data Science (DS) requires strong mathematical foundations to both understand clearly, articulate tersely, and innovate rigorously the plethora of algorithms and frameworks that have evolved in the vast field of AI/DS over the last century. This course revisits and rebuilds some of these mathematical foundations for a rigorous study, intuitive understanding, formal communication of algorithms, and express future in AI/DS.

“Data is the new oil”. Data plays a central role in driving decisions in the products and services we use across social and industrial verticals. Managing - capturing, cleaning, storing, querying, and processing - data is a foundational skill for any data scientist.

This foundation course builds an appreciation for the variety, velocity, volume, and veracity of modern data and introduces the various types of databases and best practices for managing large volumes and diverse variety of data with modern data warehousing frameworks.

Today’s social, industrial and government enterprises generate tremendous amounts of data on a daily basis. Ability to process large volumes of data requires sophisticated distributed and parallel data processing frameworks, tools, and cloud infrastructures. Handling high volumes of data is an essential skill for a data scientist.

This foundation course introduces the principles, tools, skills and a mindset for big-data thinking for building efficient big-data pipelines - from simple big queries to building complex multi-stage data pipelines on peta-bytes of data.

Data contains a wealth of insights about the process that generated it. Finding deep insights from data by asking it the right questions, interpreting the answer to those questions correctly, and letting the data speak to us is an essential skill for a data scientist.

This foundation course will explore the art, process, and algorithms to discover deep structures in data, build meaningful and real-time scorecards & dashboards for all business and operational stakeholders, find actionable insights from data addressing both first order (I know what I don't know) and second order (I don't know what I don't know) of ignorance, and tell stories about the “why” from the “what” with data.

Ability to discover structure or learn causality from past data and generalize this learning to interpret or make predictions on current data is an essential property of an intelligent system. The field of Machine Learning (ML), that focuses on replicating this aspect of intelligence into machines, has shown tremendous value across use-cases, data types, and business verticals.

This core course covers the theoretical, algorithmic, practical, process, & creative aspects of the vast field of Machine Learning.

Language can be considered as our “sixth sense” that has literally made humans the dominant species on the planet and made the modern human civilization possible. It is the primary mechanism used throughout human history to communicate and share knowledge both across society and across generations. One of the holy grails of AI is to build Machines that can “understand” (analyze), “generate” (synthesize), and “translate” language the way humans do so that we can communicate with machines in our natural languages the way we communicate with each other.

This core course in Natural Language Processing (NLP) covers the broad algorithmic, statistical, and linguistic foundations of natural language understanding and natural language generation, introduces the end-to-end process of building NLP capabilities, and explores the wide variety of applications of NLP across social and industrial verticals.

Speech is the most ubiquitous interface of communication among humans. It is powered by our vocal tract that generates speech and our auditory system that understands speech. An important milestone for AI is to replicate and master the process of speech “understanding” (analysis) and speech “generation” (synthesis) by machines. As a result, the next generation interfaces between humans and machines can become “speech-first”, leading to a new era of “inclusive AI” where the next Billion who are not as tech savvy can also “speak to the physical-digital systems” and get their questions answered, content delivered, tasks done, and problems solved.

This core course on Speech Processing covers the anatomical, acoustic, phonetic, linguistic, and modeling aspects of both speech understanding and speech generation.

Human Vision, powered by our eyes and the visual cortex - that has evolved over billions of years, is our most dominant and important of senses. One of the holy grails of AI is to build machines that can “see” the way humans do - not only the normal RGB images and videos but complex images such as X-rays, MRI’s, hyperspectral, etc. Machines that can interpret visual data can help scale and improve quality of our healthcare decision making based on visual data, building smarter and safer cities, creating visual human interfaces for search in retail, media, fashion, agriculture, and manufacturing industries, enable robots, drones, and autonomous vehicles to navigate safely, etc.

This course covers the biological, algorithmic, modeling, architectural, process, and application aspects of analyzing and interpreting the visual world around us, to build “seeing machines" across a variety of visual sensors, and enabling use-cases across industry and society verticals.

One of the promises of AI/DS is to improve the efficiency, reduce cost, and increase robustness of operations, increase utilization of resources, and improve customer experience in the modern digital-physical ecosystems. “Self-optimizing Operations” via “end-to-end optimizations” is the hallmark of today's “AI-first operations business”. A wide variety of optimization algorithms and frameworks have evolved in the last century for a variety of use-cases.

This core course explores the theoretical and algorithmic foundations of various optimization frameworks and their applications across the industry. More importantly it develops the art of formulating business problems into optimization problems.

Unstructured and semantically low-level data such as images, videos, audio, and text is becoming increasingly common in enterprises and society as people and systems generate and consume such data in both public and private settings. Human systems of visual, auditory, and linguistic perception and generation have mastered the art of processing such data in the brain. Modern Deep learning systems try to mimic a similar multi-layered architecture to interpret this low level data (e.g. pixels, signals, words) into high level semantics.

This course explores the biological, computational, theoretical, architectural, and modelling aspects of deep learning models that come in a variety of architectures, on different data types, and used for a variety of applications.

“With great power comes great responsibility”. AI is emerging as a tool for fundamental and disruptive transformation of human society. If not used ethically and responsibly, it has the potential of doing more harm than good to individuals, societies, and the human civilization as a whole. Strong regulations are emerging across the globe to ensure that AI is used for the right purpose, in the right way, and with the right governance.

In this core course we will understand the emerging guiding principles of Responsible AI, the state-of-the-art algorithms used to detect and correct any intentional or unintentional transgressions by AI systems w.r.t. these guiding principles, and the emerging landscape of regulations for AI across the world. This will enable emerging data scientists to “consciously and proactively” incorporate these principles and regulations into their model building from the start and not as an afterthought.

The last decade has seen a resurgence of deep neural networks as the machine learning tool of choice for many application areas including computer vision, reinforcement learning, and natural language processing. This course offers a holistic and hands-on introduction to deep networks, their many varieties and applications, as well as the algorithms used to train them. In this course we discuss the foundations of Feedforward Networks, Convolutional Networks, and Recurrent Networks, as well as their usage within popular Reinforcement Learning frameworks. Using real datasets and basic Python libraries for data manipulation, vector/matrix algebra, and automatic differentiation, students will code up - from scratch - popular examples of each such network architecture and apply them to applications in natural language processing, computer vision, speech recognition, and automatic control.

Human-computer Interaction (HCI) studies the relationship between computer technology and people. HCI is a multidisciplinary approach to observing how to best design computers and other forms of information technology for our everyday user needs and enjoyment. It borrows methodologies and techniques from Computer Science, Human-factors Engineering, and Cognitive Science to make our interaction with technology as beneficial as possible. HCI is important to learn because it’s designed to close the gap between user and device to maximize efficiency, with some even using it to improve the happiness and satisfaction of users. This is applied through user-interface design, accessibility (e.g. screen readers), keyboards, and other tools that help us use technology more easily and effectively. This course will help students to learn to implement HCI concepts into a variety of applications specifically web development.

Data structure and algorithms are two of the most important aspects of computer science. Data structures allow us to organize and store data, while algorithms allow us to process that data in a meaningful way. Learning data structure and algorithms will help you become a better programmer for artificial Intelligence, operating systems, graphics, and any other fields. You will be able to write code that is more efficient and more reliable. You will also be able to solve problems more quickly and more effectively.

The course cover development techniques including basic ideas of correctness and proof; Recursion. Approaches to Problem Solving, notion of abstract data type, representation of lists, stacks, queues, trees, graph, and hash tables, various algorithms, and complexity and analysis; and choosing data structures.

This advanced elective course builds on the foundations of the core Machine Learning course and introduces advanced topics in Machine Learning.

This advanced course on statistics explores some deeper statistical constructs used for more sophisticated number crunching, find deeper patterns in data, and analyzing it to find complex structures.

This advanced elective course explores the evolution of Natural Language Processing from “traditional” approaches to “deep learning” architectures specifically designed for text data and to solve specific NLP tasks with this data.

This advanced elective course explores the evolution of Computer Vision from “traditional” approaches to “deep Learning” architectures specifically designed for visual data and to solve a variety of vision tasks with image and video data.

This advanced course explores the evolution of speech processing frameworks from “traditional” to “deep learning” architectures specifically designed for speech data and to solve core problems such as speech recognition, speech synthesis, speaker identification, etc.

A large class of data in different verticals such as financial and economic data, wearables data, geological data, IoT sensor data, or alarms and alerts data come in the form of one or more, synchronous or asynchronous Time-Series. Analyzing such data gives us the ability to understand and improve the generative process, find temporal and causal nuances in the process that generates it, and build prediction, forecasting, and outlier models from such data. This core course will explore various types of time-series analysis algorithms with varying degrees of complexity and suited to different types of applications and use-cases.

Today’s consumers in any domain have many choices - content to consume, people to connect with, products to buy, places to travel, courses to take, jobs to apply to, etc. Sophisticated Recommendation algorithms and frameworks have evolved over the last several decades that can sift through the past choices a particular consumer and similar consumers made, build a deep profile of the consumer along various aspects, and use this profile to suggest the most likely items from a very large pool of items the consumer might like to consume next.

This core course will explore theoretical, algorithmic, and practical aspects of recommendation engines and explore their applications across various domains.

This advanced elective course builds on the foundations of the core Reinforcement Learning course and introduces advanced topics in Reinforcement Learning.

AR/VR (Augmented Reality/ Virtual Reality) Lab particularly focuses on interaction technologies including virtual, augmented and mixed reality as well as mobile computing, epigenetic and evolutionary robotics, and haptic communication.

This lab provides exposure to students to the complexity, breadth, and depth of how vision works and how to make machines “see the way humans do” in different areas including healthcare, agriculture, retail, autonomous vehicles, drones, manufacturing, security, traffic management, remote sensing, defense applications, and so on.

Media Studio is designed with state of the art equipment used in newsrooms for students to gain extensive practice in the technology and process of content creation. Students will get hands-on training in the audio, video, and social construction of media and the associated best practices in audience research.

Production Control Lab will complement the media studio to mimic the environment and technological standards of the live creation of digital media and longer feature programming. Students will become skilled at lighting design, reading from a teleprompter, managing multiple audio and video inputs, in-ear communication and monitoring, robotic and person-operated cameras, pacing and voice, and collaborative construction of media at this lab.

Video Editing Bay includes state-of-the-art audio equipment and is also be outfitted for HD, Ultra-HD, and 4K video editing including Adobe Creative Suite software, a playback interface, a color console, and a professional reference monitor. In these rooms, students will be able to refine and edit video shorts, news pieces, feature-length stories, documentaries, and AR/VR content.
 

Audio Editing Bay comprises acoustically treated rooms that will be equipped with sound-mixing consoles, associated software, surround sound, and amplification. These rooms will allow students to record and edit podcasts, and voice-overs for films and AR/VR applications.