AI / Data Science Education in India

In a first-of-its-kind study in India, IUSSTF partnered with itihaasa Research and Digital, Association for Computing Machinery (ACM) India, and National Programme on Technology Enhanced Learning (NPTEL) to conduct a survey of Indian academic institutions to (a) assess the current state of AI, Data Science, and AI-related curriculum, (2) understand the underlying challenges, and (3) identify infrastructure and resource needs, including faculty recruitment. The survey was sent to ACM Chapter members, NPTEL Local Chapters, and IUSSTF grantee institutions. If an institution had multiple academic units offering degrees/ programs, each academic unit was asked to only provide information related to their specific unit.

Many institutions offer multiple programs and therefore, the number of programs will not equal the number of respondents. Response rates may also vary because some institutions provided only partial responses.

While we provided a definition of ‘Artificial Intelligence’ and ‘Data Science’, (Annexure I), respondents were asked to use the definition that was most appropriate for their specific program/ degree. The survey (Annexure I) was divided into three sections: Section A addressed Program Offerings, Section B focused on Curriculum and Instruction, and Section C addressed Faculty, Infrastructure, Resources, and Collaborations.

A total of 113 responses (105 unique institutions) were received from a wide range of public and private institutions across India. The list of participating institutions who consented to have their names published is provided in Annexure II. We acknowledge that this was an exploratory study and some of the inferences/ conclusions may not be generalizable. As a follow-up, a more extensive study involving a larger number of Indian institutions should be conducted.

3.1 Program Offerings

Among all respondents, 62% (69) of institutions indicated they are offering programs in AI, Data Science, or related areas, with only 10% reporting no plans to offer such programs in the future (Figure 1).

Figures 2 and 3 summarize the degree programs offered at the Bachelor’s, Master’s, and Ph.D. levels. Of the 69 respondents, 80% (55) offered Bachelor’s programs, 57 % (39) offered Master’s programs, and 58% (40) offered Ph.D. programs.

Respondents were asked to describe the types of AI or Data Science Programs/ Degrees offered by their Academic Unit (Department, School, Center, College). There were four possible options at the Bachelor’s, Master’s, and Ph.D. levels:

» Stand-alone AI Program

» Stand-alone Data Science or Analytics Program

» Computer Science program with a specialisation in AI or Data Science

» Programs in other disciplines with a specialisation in AI or Data Science

In addition to formal degrees, respondents were also asked if they offered any one or more of the following programs:

» Online program/degree in AI, Data Science/Analytics at the Master’s level

» Online program/degree in AI, Data Science/Analytics at the Bachelor’s level

» Online Training Courses/Workshops in AI and Data Science

» Training Programs in AI and Data Science

Figure 4 provides the percentage of institutions offering each of the different programs. One surprising finding is that 42% of respondents offered Bachelor’s stand-alone AI programs or Bachelor’s programs in Computer Science with a specialisation in AI or Data Science. Conventional wisdom would suggest that a larger number of institutions would offer programs in AI and Data Science at the Master’s level. The fact that such a specialisation is seen at the undergraduate level may be a reflection of the demand from students as well as market demand from industry.

Another reflection of the market demand for AI and Data Science skills is the fact that nearly a third of the respondents offer training courses / workshops and certifications – both in person (29%) and online (30%). The high prevalence of online training courses could also be a consequence of the Covid pandemic.

We are also beginning to see the emergence of online degree programs in AI and Data Science at both the Bachelor’s (4% of respondents) and Master’s level (14% of respondents). While the numbers are small, we see this as an encouraging trend given India’s push toward enhancing the Gross Enrolment Ratio (GER) in quality higher education.

Institutions also provided information about programs in AI, Data Science, or related fields not covered by the categories listed in the survey. These include:

» BTech (Data Science)

» BTech in Information Technology, BTech in Information Technology with Specialization in Business Informatics

» BTech in AI and Data Science. Minor in AI and ML for engineering students

» BTech In Math and Computing

» BTech Honors and Minors in AI &ML, AI&DS

» MTech in Information Technology with Specialization in (1) Data Engineering, (2) Human Computer Interaction, (3) Machine Learning and Intelligent Systems, (4) Robotics and Machine Intelligence

» MTech Data Analytics and Decision Sciences

» Interdisciplinary Dual Degree Program in AI ML and Applications

» Executive MTech in AI

» Quantitative Economics (MSQE) and Quality, Reliability & Operations Research (M. Tech. in QROR), MS in Quality Management Science

» PhD in Machine Learning

3.1.1 Academic Units Housing AI and Data Science Programs

Figures 5-7 provide information about the academic units housing different programs.Conventional wisdom would suggest that AI and Data Science programs are housed in a Computer Science department, and the data for Bachelor’s, Master’s and PhD programs certainly indicate this to be the case. But a surprising finding of the study is the existence of a unit dedicated to AI in many institutions. Given the small number of respondents to some of these questions, the trend, rather than the actual percentages, is certainly interesting.

The creation of a separate AI department may be a recognition of the fact that AI has permeated every discipline of knowledge, be it engineering, medicine or humanities, and the increasing demand for such programs. The study captures some of this diversity with institutions reporting programs in disciplines such as agricultural sciences, biological sciences, and medicine that include a specialisation in AI or Data Science.

This is certainly encouraging and a sign that Indian academic institutions are responding to the global trends. The timelines involved in the creation of India’s first CS department in the 1960s provide an interesting contrast. At IIT Kanpur, computer science courses were initially offered by electrical engineering faculty as there was no separate CS department. It took a couple of years before MTech students in the electrical engineering department were allowed to choose “computer science” as their specialisation. It took another decade before CS was available as an area of specialisation for the under-graduate program.

A high-level analysis of the institutions that participated in our survey reveals that several private engineering institutions and the newer IITs / NITs / IIITs have embraced the idea of offering stand-alone AI programs at the Bachelor’s level and have also established separate departments dedicated to AI. This may be an effective strategy to attract students and recruit top faculty members to their institutions. We also find that the older and more established institutions like IISc and the five original IITs have developed stand-alone AI programs at the Master’s level.

3.2 Curriculum and Instruction

In this section of the survey, we provided a list of courses and asked respondents to indicate whether the course was (a) Required or (b) Elective or (c) Not Offered. Figures 8 and 9 indicate the percentages in each of the three categories for Bachelor’s level and Master’s/Ph.D. level courses respectively. Respondents could choose to skip over an item so the number of institutions may be different for different responses.

From the graph, we see that the following courses are not offered by 25% or more of the departments as part of their bachelor’s program: (i) Philosophy of AI, (ii) AI & Brain sciences, (iii) Multi-Agent Systems, (iv) Human-Computer Interaction, (v) Speech Processing , (vi) Robotics & Automation, (vii) Ethics (privacy, fairness, explainability), and (ix) Text Mining. The same patterns persist at the Master’s level. Except for Text Mining, 25% (or more) of the departments do not offer the remaining eight courses listed above. While some of these topics are quite specialized, any undergraduate program must include coursework in Ethics addressing critical issues related to privacy, fairness, explainability, and trust.

One of the major challenges in offering these specialized courses is the lack of faculty with expertise in these areas. One solution to address this gap is to leverage online courses available on platforms such as NPTEL or other third-party education platforms such as Coursera / EdX / Upgrad.

Institutions were asked to list other courses offered as part of their AI and Data Science programs. The list includes:

» Bayesian data analysis, financial data analysis, risk management, survival analysis

» Introduction to smart sensing, information retrieval, fundamentals of sensors and transducers, pattern recognition, introduction to R and Python, data curation, data warehousing, image processing, business analytics

» Soft computing, visual recognition, convex optimization, advanced data analytics, computational astrophysics

» Robotic perception, data driven control

» Database management, advanced computer vision, machine learning in bioinformatics

» Computational finance, Computational psychology, algorithms for big data, information retrieval, computational biology and bioinformatics

» Fundamentals of data science

» Computational data science, time series and survival analysis, tools for big data computing

» Cloud architecture, reinforcement learning, game theory in AI, GPU computing, evolutionary algorithms, data visualization

» Information systems, distributed computing, computer integrated manufacturing

» AI for cyber security

» Reinforcement learning, signals and systems, image processing

In addition to coursework, we also wanted to understand what additional learning and training opportunities are available to students to prepare them for careers in AI and Data Science. Figure 10 shows the percentage of responses in each of the three categories: Required, Optional, Not Offered.

While the data reveal that a majority of departments require some form of research as part of the training, this is not surprising given that 58% of institutions (see Figure 2) that responded to the survey offer a Ph. D. program. If we consider only the 25 institutions that do not offer a Ph.D. program, the figure drops to 44%. What was also surprising and counter intuitive was the drop (64%) in the requirement for hands-on projects in this category. While the sample size is not sufficient to draw inferences, this may explain some of the findings related to the lack of access to quality datasets. The data indicate that 71% of institutions offer online courses including MOOCs. While this may be a consequence of the pandemic or a dearth of faculty, it would be interesting to see if educational institutions continue to embrace online instruction and flipped classrooms. Two areas that need to be addressed include increasing interactions with industry and a more interdisciplinary curriculum.

3.3. Faculty, Infrastructure, Resources, and Collaborations

This section of the survey addressed the state of resources (human capital and technology infrastructure) available in institutes for teaching AI and Data Science, and factors that impact the institution’s ability to offer these types of programs.

Faculty Strength and Expertise

Respondents were asked to indicate the faculty strength of their respective academic units. Form the table we see that a third of the departments have fewer than 10 faculty members.

Department Size	Percentage
01 - 10	33
11 - 20	20
21 - 30	12
31 and above	35

We also asked respondents to estimate the total number of faculty members in their academic unit with core experience / expertise and training in AI and Data Science. Only 29% estimated that their academic unit had a high proportion (> 50%) of faculty members with expertise in these areas. This dearth of trained faculty with expertise in these areas is certainly a major concern. We need to significantly increase the numbers of AI / DS faculty members in India to address the demands of the AI Workforce.

To assess the extent to which the programs are inter-disciplinary, we asked respondents to indicate whether or not faculty members from other Academic Units were involved with their specific AI or Data Science degree/ program (Figure 11).

As is to be expected, Faculty from the Departments of Mathematics/ Statistics, Computer Science, Electrical / Electronic Engineering, and Artificial Intelligence are heavily involved with the AI and Data Science programs offered at their institutions. It is heartening to note the involvement of other engineering (35%) departments as well as departments of business/ management (28%), biological sciences (20%), and social sciences (17%).

While the participation from departments like medicine (4%) and agricultural science (4%) is low, this is not surprising as many of the institutions that participated in the survey focus on engineering/ technical education.

3.3.1 Factors affecting AI / DS programs

We asked respondents to assess the extent to which (greatly, to some extent, not at all) each of the following factors affected/ impacted their AI / DS programs. Figure 12 displays the net score (greatly + to some extent – not at all) for each factor.

The lack of availability of quality datasets and access to computing infrastructure are clearly major challenges. High quality, annotated, benchmarked, domain-specific datasets are required to train AI models and systems. While there are a number of Government Agencies in India that collect data, issues of interoperability and provenance will need to be addressed to ensure the availability of large datasets for research and training. Industry can help in this regard by providing real-world data for student projects. When it comes to high-end computing infrastructure, institutions need access to Graphics Processing Units (GPUs) and specialized hardware such as Tensor Processing Units (TPUs), Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs). Access to cloud infrastructure may address some of these issues.

External collaborations, both with industry and global universities and research organizations, is another challenge faced by many institutions. The need for deep collaboration with industry is clear - ensuring the availability of quality datasets, access to internships, and input on curriculum and training to ensure students have the knowledge and skills to enter the AI workforce. International and Public-Private partnerships are critical to the development of a diverse, globally-engaged, technology-abled workforce. The USIAI can serve as a platform to facilitate collaboration between Indian and U.S. academic institutions to address the challenges associated with developing a diverse, robust AI workforce.

The same question was posed to Institutions that are planning to offer Programs in AI or Data Science in the near future. While there are many similarities, Collaboration with industry and Lack of Faculty with expertise in AI and Data Science are the biggest challenges (Figure 13).

3.4 Career Paths/ Aspirations of Students in AI and Data Science Programs

Respondents were asked to describe the career paths/ aspirations of students in their AI and Data Science Programs by ranking the four categories on a scale of 1 (Most preferred) to 4 (Least preferred) (Figure 14)

It appears that the overwhelming preference for students (across undergraduate and graduate AI/DS programs) is to join the Tech industry. Careers in academic / research streams seem to be lower in priority. Funding agencies and academic institutions can make research positions more attractive through fellowship and mentoring programs. The extremely low preference for startup / entrepreneurial careers may be an indication of the inherent risks, lack of exposure to incubators, and most importantly capital. Students may prefer to join an existing startup (which may be counted under Industry (Tech) careers).

The survey, while exploratory, provides an interesting window into the AI education and training landscape. While the number of programs is growing rapidly, we need to also evaluate the quality of these programs in terms of research productivity, and knowledge and skills. itihaasa’s landscape study of AI research in India also found that the quality of the Indian AI/DS programs has to be greatly enhanced in order to meet the requirements of industry and research. Accrediting bodies, professional societies, and industry associations can play a role in setting guidelines and standards for new programs.