Search This Blog

Monday, February 20, 2017

IIIT Delhi: The Journey towards Excellence

I have been invited to talk about achieving excellence in universities. The best way to illustrate your points is by way of an example. Indraprastha Institute of Information Technology Delhi (IIIT-Delhi) has had a dream start. It is widely recognized in academic circles as a place which provides excellent education and generates excellent research output, comparing with the best institutions in the country. How did it happen. Are there lessons to be learnt here. Well, read on. (Statutory warning: There are far too many stories to tell. It is long, and is written for an audience who is genuinely interested in knowing the magic that we are doing in a small campus in South Delhi.)

The Act of Delhi Legislature

The journey of a new university starts with an Act, either by parliament or by a state legislature. (Illegal things do happen. Many IITs started without an Act. But we will ignore that.) IIIT Delhi was established through an act of Delhi legislature in 2008. This is arguably the best act in terms of autonomy that it bestows upon the Institute and its board. I do not know who all were involved in drafting the legislation, so I can't thank them in this article, but they seem to have studied other successful models of university administration (like IITs), learnt from the problems they face, tweaked those models for a state university, and come up with something which is absolutely remarkable.

As an aside, the first university act in India was passed in 1857, setting up the University of Calcutta. It was an act that took care of all colonial interests. You will notice that the administrative structure of the university was not the same as those of successful universities in England at that time. The goal of the British was not to replicate excellence here. Far from it. The goal was to ensure that the university remains tightly controlled by the colonial powers. And hence, they had a Vice Chancellor who would be the supreme leader with no checks on his power. They could simply have a white person who understood English interests as the VC who would ensure that the university campus does not become a hotbed of nationalism. 160 years later, we continue to have the same colonial model for administering our universities even though very successful models exist in the country both in the government sector (IITs) and the private sector (like BITS). It was extremely thoughtful of the Delhi legislature to set up IIIT-Delhi without following the colonial model.

Leadership (Founding Director)

The next step in setting up of a university is to find passionate and visionary leadership. All too often, one finds that a university has been set up and even admissions have been announced with no academic leadership in place. Directors or VCs are appointed long after the programs have started. Who decides what programs to start. Who decides what curriculum would be followed in those programs. Who decides who will teach in those programs. Temporary and part-time administrators may sometimes do wonders but those are the exceptions that prove the rule. And the beginning of a university is a very sensitive time. The culture and the public image that gets created in the beginning will take a long time to change.

In this regard, once again, IIIT-Delhi was very fortunate to have Pankaj Jalote as its founding (and current) Director, and Kiran Karnik as the Chairman of its board.

The leadership can really make or mar a new institute. The biggest pressure on the leadership is to ensure that classes take place. For that, you need faculty. And I have been associated with a large number of great universities as well as a number of not so great universities. The difference between the two is simple. One set of universities went for distress recruitment, while the others ensured that whatever the pressures may be, we will only recruit the top quality faculty. To do the latter, the leader must be able to invite good teachers from any good institute nearby as visiting faculty. Someone who is well respected in academia can do this more easily. Also, you need to ensure that while there may not be any faculty recruitment in the first few months, soon there will be great faculty getting attracted to this new place. Again, having a great leader helps. And, of course, you don't find faculty based on an advertisement.

Faculty Recruitment 

So, the third great thing (after the Act and the leadership) that happened to IIIT-Delhi was recruitment of quality faculty. It is a common refrain that there is a serious shortage of faculty. Yes, there is. The way to look at faculty recruitment is this. We need 20 good faculty members. Are there 20 good PhDs in the job market. The answer is, obviously, yes. Now, how do we get those 20 to join me rather than join an IIT or another good institute. There may be a serious shortage of faculty in the country, but we want only 20, and there are more than 20 good PhDs in the job market.

First of all, you need to inform all of them about your existence. Advertise in all the right places. No, I am not talking about Times of India or Hindustan Times. PhD students around the world don't read faculty advertisements there. May be CACM would be better, or some IEEE publication, or various mailing lists. Perhaps you need to inform all your contacts in all good universities and ask them to forward your mail to PhD students.

Then you need to convince them that they would have a great future if they join this new place. Now, a Director, who himself is very well decorated, will have an easier time convincing. If Pankaj Jalote can join this place, why not you. But, of course, communication is important. In a face to face meeting, the potential faculty members can better make out the passion with which this Director is working. So you need to travel to those universities, and may be invite them to your campus or whatever you have in the name of temporary facility. Speed is important. Most universities would take weeks to even acknowledge an application, and would take months to make an offer. If you can acknowledge within hours, and make an offer within days of meeting/seminar, etc., you can win the game. And once a few great people join, you would be noticed. These young recruits would do the recruitment for you.

Pankaj never did any distress recruitment. Every faculty member recruited was on merit, and could have easily joined an IIT or another top institution in the country. This really is the biggest reason for IIIT Delhi's success. You recruit a few average faculty members. They become the benchmark for others. And then they demand promotions at weaker performance levels since they bailed you out when you needed them most. And once you promote some people for institution building, you are on a downward slope. Thankfully, this did not happen at IIIT-Delhi.

Tenure System for Faculty

A huge gamble was taken at the very beginning and that was to implement a tenure system for faculty. Everyone was shocked. In a country where IITs offered "permanent government jobs," why would anyone join on a contract for 5-6 years. You are new. No one knows about you and then you will make an offer that is not as good as other well established players. Well, it was precisely because IIIT-Delhi was new that it had to offer a tenure. Just like employees need job security, the university needs security from a poor employee. An old large institution can afford a few bad apples, but a young small university can not. And it can be no one's case that there would never be hiring mistakes. It was thought that most young PhDs who have done exceptionally well in research till then would really not mind being on contract for a few years, if they really believed in the future of IIIT-Delhi. And Prof. Jalote was proven right. Several young PhDs from top places in India and abroad joined and the rest, as they say, is history.

If faculty was on contract, it was so much easier to have all staff on contract as well.

Annual Reviews

Yearly reviews of all employees - faculty as well as non-teaching staff - was another great idea that IIIT-Delhi implemented. Every faculty member submits a detailed form giving out all sorts of information on the research (papers, patents, consultancy, projects, software developed, PhDs guided, and so on), teaching (number of courses taught, which courses, how many students, what was student feedback in each of them, new courses, writing of books and any other instructional material, BTPs, MTech theses, and so on), and on service (admin work in the Institutes, committees, conferences/workshops organized, professional work outside the institute, etc.). The information is analyzed by experts outside IIIT-D. Individual feedback is given to each faculty by the Director. There is also a possibility of faculty member rebutting the analysis which is again sent back to the experts and they may want to change their feedback. A small bit of carrot. Those who do extremely well in annual review get a slightly higher support for professional expenses.

The staff feedback is 360 degrees. Not just the officer above and reportees give feedback, all those whom the office was supposed to serve are also asked to give feedback on a variety of questions. All this is consolidated by a small committee and then individual feedback is given to each employee. Come promotion time, there won't be any surprises. One can even identify the employees who are doing so well that one can think of an early promotion for them.

PhD Recruitment

So, we got the best faculty. But how do you make them productive. Well, need a strong PhD program. Why would any good student join an unknown place for PhD. They would because thankfully, most good institutes do not care and have placed a lot of technical restrictions. In 2008, pretty much every IIT claimed that they recruited fresh BTechs without GATE score in their PhD program. But the reality was that such numbers were close to 0 in every place. So the faculty went to nearby good colleges - NSIT, DTU, MNIT, LNMIIT, IIIT, JIIT and so on, and talked the graduating students. Asked them to apply right there, took an interview, and made offers. Very soon, the quality of PhD students, if you can judge that on the basis of pedigree alone, was better than most IITs. But, the quality of PhD program should be judged not by the pedigree of the student, but the publications during the PhD program and the quality of places that they join after completing the PhD. And the first 10 PhDs that have already graduated have proven that indeed today, IIIT Delhi has among the best PhD programs in the country.

The PhD students were treated well. They have the first right to the hostel rooms. They get a free laptop when they join. Each PhD student is given one's own space with a PC. They are given a budget for conference travel, both for within India and outside. There is a contingency budget for small expenses. There is also provision for many of them to go abroad to a good research lab for a semester. But they have to follow a certain discipline. There is an annual review. You need to convince a committee that you did work hard. Sometimes results will not show that. But if you did not even put in effort, you are probably not at the right place. There are deadlines for various stages. They are all flexible, but you will get reminders, and your supervisor will get reminders. We want each of them to graduate in 4-5 years. Yes, it is not easy to predict things in research, but in most cases when students take more time, it is because of lack of time management and not because they tried many things and failed. Let us see how successful we can be with this goal.

Teaching and Learning

A major challenge for all universities is to excel in both teaching and research. There is a common belief in many top institutions in India that teaching courses lead to reduced quality of research. (You would hear this all the time: IISc professors are able to do better research because they teach less.) First of all, the Director believed that what is important is that students learn and are ready to go to the next stage of the career - whether a job or higher education - well prepared to take any challenge. And to ensure that learning takes place, we not only want good teaching, but more importantly processes and pedagogy designed to focus on learning.

So, a couple of teaching learning workshops were held where experts from around the world came and talked about techniques which can help you focus on learning. Just imagine the impact on young faculty members who had never even taught. Even later on, we continued with the system of sharing best practices in teaching in a faculty meeting.

Curriculum

The curriculum was decided based on several beliefs. One, it is better to teach 5 courses in a semester with a lot of assignments, projects, labs, etc., rather than teaching 6 courses with much less engagement. We told every potential student even before the admission - we expect you to work 10 hours a week on a 4-credit course. So in a typical semester, about 50 hours a week (including about 20 hours of lecture/tutorial/labs). Some students/parents got scared. Other engineering colleges talk about 30 hours a week. We were completely OK with that. Learning any new skill takes a lot of effort and we were not in the business of issuing degrees. We are in the business to ensuring that each and every graduate meets all the program outcomes well.

The other belief was that it is best to do engineering subjects in the first year. They have done enough science in high school. They are coming to an engineering college to learn engineering and are greatly disappointed when they are taught the same things that they studied for JEE preparation. It also meant that we could complete all the core courses by the second year, enabling them to do really great internships. They could then choose their electives based on their interests in 3rd and 4th years. They could also do a BTech Project over a longer duration (say, 3 semester).

And of course, the under-graduate programs are for broadening of horizons. Also, many students pick up programs without really knowing what they are getting into. To solve both these problems, a large number of electives are part of the curriculum. Pretty much the entire 3rd and 4th year program consists of electives - some have to be done from humanities and social sciences, some have to be in the discipline, and a whole lot of them are just open electives - they can do anything under the sun.

To provide even more flexibility, online courses were allowed right from the first batch. So many great universities are offering so many great courses for free. Why not leverage them and enable students to learn topics for which we do not have faculty - jazz appreciation, just as an example. One may limit the number of credits earned this way. One may mark those credits as Pass/Fail and not count them towards CGPA, but allow learning beyond what our own faculty can offer.

Yet another flexibility is that of Independent Study. Many times a student is interested in learning a topic for which there is only a limited interest overall and hence a regular course can not be offered. Well, a student can study that topic with a faculty as "independent study." So a course can be offered for even one student.

To encourage under-graduate research, the curriculum not only has the option of a BTech Project, but also extra research credits can be earned under "UG Research" and "Independent Project."

With all this flexibility, each student can plan his/her own degree. One can additionally use the open elective slots to do minors.

To understand what BTech students learn, one only needs to visit IIIT-Delhi during the end of the semester. Many courses have their students put up a poster on their projects. And the kind of things that students do as course projects is simply amazing.

Bonus Marks for Admission

IIIT-Delhi is about innovating in every sphere. There is nothing routine about anything that we do here, as it should be for any university. And admissions to UG programs are no exception. From early days, it was felt that performance in a single test must not be the only way to seek admission. We must look at other achievements of students and give these achievements some weight in our admission process. This gave birth to the idea of "bonus marks." For a large set of achievements both in curricular and extra-curricular sphere, a student will be awarded certain bonus marks (a maximum of 10 percent). These marks are added to whatever exam we are using as the primary means of admission (JEE in the last few years). The achievements include NTSE scholarships, Olympiads, performance in the inter-school programming contest in Delhi, having a rating in chess, winning sports medals in national school games, getting cultural fellowships from Government of India, and so on. And being a data driven university that we are, every year, we check the average CGPA of our class and the average CGPA of those students who had received bonus marks at the time of admission, and every year we find that students with bonus marks perform better than other students.

Starting New Programs

Starting a new program is a very serious business at IIIT-Delhi, as it should be in any university. For any proposal to start a new program, we would search for the best current offerings anywhere in the world. We will have several workshops where we will invite experts in those or similar areas to debate each and every point related to the program. What should be the program outcomes of the program. What course structure would allow students to learn those outcomes. What should be the ordering of the course. What career options would the graduates have. We would also have workshops with industry to see if they will really be interested in the output of such programs. And then we go back to experts to help us design the detailed curriculum of each course, and also seek their help and support for recruiting quality faculty. I have not seen such due diligence in any other university that I have been associated with. And yet, if we think about it, this is how all new universities should decide on the programs that they will offer.

After starting with Computer Science and Engineering programs in 2008, we have later added PG program in Computational Biology, both UG and PG programs in Electronics and Communication Engineering, UG program in Computer Science and Applied Maths, and we are about to start two new UG programs: Computer Science and Design, and Information Technology and Social Science. To learn more about 

Outward Looking

In general, a new university has to be outward looking. One wouldn't have all the expertise in house. In fact, one would never have all the expertise inhouse. And therefore, IIIT-Delhi is always drawing upon the support of well wishers who want a good university to succeed. Even when we got the land from Delhi Government and had to build the campus, we organized several workshops involving experienced faculty, civil engineers, architects, and so on, to come up with the requirements and even detailed design. Same thing happened when we started planning for the second phase.

Support from well wishers

Last year, we set up an Infosys Center for Artificial Intelligence. This was one of the largest industry gift to an academic institution in India for supporting research. We are now seeking support from all sources. The strategy of the founding Director has been that we must first prove our mettle, and once we have built a name for ourselves, we should reach out to various well wishers, show them the quality of research that we have already achieved, and then ask for support.

There are more things to tell. So I will keep adding more stuff here over the next few days.

Sunday, February 19, 2017

CS education is poor because of copying

The CEO of Capgemini, Mr. Srinivas Kandula, has mentioned that 65% of IT employees are not trainable. We have all along been hearing that only 15-25% of the graduates were trainable and that the IT industry was only hiring from this group. Now, we are being told that even within this lot, 65% are really not trainable, and were presumably hired since there were really no option, and perhaps there was enough very low quality work to be done there. And now that the growth rate of the industry is coming down, and there is uncertainty in their largest market, it is perhaps better to recruit freshers for such low quality jobs than to pay for the experienced hands.

I do not know how much to believe Mr. Kandula. But I have always been amazed at the Indian software industry. That it can grow so fast and become so big despite the absolutely abysmal quality of education in our colleges. I have always believed that our engineering graduates have learnt nothing in most colleges, but in any large sample of population, there would be those who are naturally gifted, and who can be trained for anything, and my belief has been that IT industry is able to recruit people with minimal knowledge and train them for whatever they want them to do.

But apparently not. What Mr. Kandula appears to be saying is that a whole lot of people hired by IT industry have not picked up the right skills even after the training. Till now, when the going was good, both in terms of there being significant amount of low quality work, as well as in terms of good revenue that could be charged for that low quality work, it was ok to carry all those people on the roll. But increasingly, that is going to become a challenge.

Recently, I was on a selection committee of a government department to recruit programmers. The salary was very decent, and many people after several years of experience in industry had applied. What is the largest program in terms of lines of code that you have written in your entire career including college education. Typical answer: 100 lines. Can you write a program for binary search: NO. Can you write a program for just traversing a linear linked list (no trees): NO. Can you reverse an array without using another array: NO. Can you just exchange the values of two variables given a temporary variable: Mostly, NO. These are all the programs we ask our first semester students who have never programmed before. Have you ever been to sites like codechef: NO. Have you ever contributed to any open source project: NO.

I have the same experience when I sit in MTech interviews. And remember, we are only interviewing those who are among the top 1-2% in GATE.

Why is it that graduates of CS programs after 4 years know pretty much nothing. (Another data point: Once, GATE office shared with me detailed data about the previous years' paper. The shocking part was that the median GATE score was 0. This was more than 10 years ago, but I don't know if there has been any improvement in this period. More than 50% of the students who gave GATE had received 0 marks or less.)

A lot of people have talked about poor quality curriculum, poor quality faculty, poor infrastructure, poor school education, and so on. I disagree. All this can not lead to such abysmal quality of graduates. And all these explanations are only to take away the blame from students themselves. There is a much simpler explanation for this: Copying in our colleges, besides laziness.

How did a student pass the course on programming, if s/he never wrote programs. One can see that there are multiple courses in a typical CS curriculum that requires a whole lot of programming. Very simple answer is that only a few students write code, and everyone else copies. And everyone gets close to 100% internal marks.

Programming is the most basic skill for anyone working as a programmer/developer/software engineer, etc. If a graduate can write a 200-line Java program and a 20-line Python program, they can not only find a job, but also climb a few steps in industry. (Of course, if you want better jobs, and better careers, you will need to learn a lot more.)

Is faculty so poor that they can't even teach just 3-4 courses well (out of 40-50 that a student will do). Even if that were the case, there is enough online material to learn programming, data structures, algorithms, etc. Colleges that want their graduates to get jobs can do something very trivial. In just one course every semester, ensure that students do not copy. If they do not submit original programs, they fail. Checking copying is very simple these days.

Why don't colleges do such a simple thing. They can immediately improve the quality of graduates without any investment. But colleges should do it at their own risk. I know of one college which tried this. Every single glass in all buildings were broken by the angry students.

Wednesday, November 9, 2016

Use unaccounted wealth to improve education

Another surgical strike happened yesterday. The Government of India has declared the Rs. 500 and Rs. 1000 notes will no longer be legal tender in the country. This is huge amount of money. Different newspapers are giving different figures, but they vary from Rs. 6 lakh crores to Rs. 9 lakh crores (6-9 trillion rupees). Of course, people and organizations who are holding this cash can deposit this money in the bank over the next 50 days, but if the money was unaccounted for, it could lead to questioning by Income Tax authorities. Most people are estimating that a significant part of this cash is unaccounted for. Of course, over the next 50 days, the financial experts will try to find ways to deposit the money in bank or get it exchanged with valid currency, but it is pretty obvious that a very large amount, in trillions of rupees, will be worthless very soon.

Will burning these notes be the only solution?

I have a suggestion. What if the government allows these people to "donate" these notes to educational institutions (who have been given tax exempt status for donations) and allow educational institutions to deposit any amount of such donations in the bank. These people who donate money can then be given a receipt which they can use to claim tax deduction.

If someone has 1 crore worth of unaccounted for currency notes, he will have the following options:

1. Give this money to an educational institute anonymously. He may feel nice that his ill gotten wealth is helping the nation in some way.
2. Give this money to an educational institute, get a receipt, show that receipt in his income tax return and get a tax saving of 30 percent (or 15 percent, depending on the educational institute). The person still loses 70 percent of the money, but again, he may feel nice that instead of destroying those notes, they are of some use in nation building.

Depending on whether someone is willing to show that money or not, they can chose option 1 or option 2.

Of course, one could consider other priorities and allow a broader set of organizations to whom such donations can be made. Since in some cases, this can become a conduit to exchange the old notes with new (you donate, and then you get a contract or some other goodie), government may decide that only government institutions will be permitted to accept such donations.

Just imagine what it can do to education.