[Techie Tuesdays] Meet the chief architect of Aadhaar, Pramod Varma
When Pramod Varma took on the challenge to architect the world's largest identity project (almost 10 times bigger than the largest existing identity system), he didn't have much idea of what he was up to. But driven by instinctive decision-making and being a believer in people's capabilities, Pramod and his bets on technology have made Aadhaar one of the most robust and scalable systems, achieving the goal of inclusion like none other.
It’s 7:30 am, and I’m 15 minutes early for my scheduled meeting with Pramod Varma at his EkStep office. But when I climbed the stairs to the office, I was hardly expecting to see Pramod well-settled with his work and having his breakfast alongside. He gave me a tour of the office and requested some coffee before we settled in for the interview.
One meeting with Pramod is enough to convince you as to why he is the Chief Architect of Aadhaar and the architect of India Stack. An interview that was meant to last for an hour went on for more than two, and I still found myself wishing for more time.
This week’s Techie Tuesdays explores the life of an exceptional techie, one who is both an artist and an architect, intent on turning the vision of providing identity to more than a billion Indians into a reality.
The royal connection
Pramod comes from the erstwhile royal family of Kerala, the same one that famous painter and artist Raja Ravi Varma belonged to. Kerala's rapid shift to communism brought many royal family members, like Pramod’s parents, into contact with the harsh realities of life. His father often used to say,
It's important to be positive and very focused on doing the right things. Results are not fully in your control, but nevertheless, if you do the right things consistently and keep your integrity intact, generally, you'll be successful.
Born in a small town near Trivandrum in 1967, Pramod grew up in Kerala and had a simple upbringing. He finished his entire schooling in Malayalam medium, with almost no exposure to English. He didn’t know about the IITs even after finishing his 12th standard. So, he went on to do his BSc Math from a local college. During this time, he saw his first computer (dot matrix) printouts when his cousin brought some from IISc (the Indian Institute of Science). Pramod also had access to an Atari device, the small programmable game machine, which his friend's elder brother had brought back from the Gulf. By 1987, he had made up his mind to go into computer science.
Pramod was also keen to get out of his small town. After finishing his BSc, he joined Hyderabad Central University to do his MSc in Applied Mathematics. Till now, he has had no mentor. He recalls,
Landing that first time in Hyderabad was very different. I never spoke Hindi or Telugu (or even English properly) before. Just landing with a small bed and a bag in the bus stand, I was trying to figure out the way to the university.
In the university, MSc Applied Mathematics students did a lot of computer science courses, like pattern recognition, meta theory, graph theory, with MTech (Computer Science) students in combined classes. Pramod topped the university in the first year and came second the next year. He says,
“But it was irrelevant because by then, I had really figured out that I wanted to go for MSc Applied Mathematics with computer science courses. That is also when I realised why math was very powerful in computer science. Understanding data structures, graph theory, and algorithms was easier.”
Later, Pramod wrote GATE and went to JNU in Delhi to pursue an MTech in Computer Science. He also completed his PhD in Graph Theory there. Because of his programming experience from his MSc days, he was a lot more confident during his MTech. It was still the early days for FORTRAN and C then.
In 1989, Pramod wrote a full-fledged vector graphic editor that could draw shapes and fill colours. This was written in PASCAL and ASSEMBLY. It gave him an exposure to why data structures matter. For example, the tool will know how to compute a shadow against a light source. After 1989, Pramod got into LISP, Prolog, and Expert systems. He recalls building a small medical expert system (primitive when compared to today’s systems, of course) as his first exposure to artificial intelligence. He even wrote a chess player program using the standard AI algorithms like tree pruning and backward tracking, but machines were primitive then, and humans always ended up beating the program.
The left brain-right brain balance
During his academic years, Pramod had kept his interest in travelling and art going strong. He was continuing with pencil drawing and water colours, which kept his left brain intact (probably because of inherited capability from the Ravi Varma clan). Engineering and logic, meanwhile, helped in developing his right brain. He even learnt to read and write Telugu when he was in Hyderabad. Pramod is a trained rock climber and mountaineer.
He realised that his three aspects were being developed simultaneously:
a) The technological part: because of his academics.
b) Extracurricular activities: consisting of mostly adventurous like skiing, rollerblading, and rock climbing.
c) Artistic side: pencil drawing, water colours, and clay modelling.
Pramod’s room on the third floor was probably the best on campus, with a lot of drawings and paintings on the wall and an equally great collection of other artwork kept all over the room (which he collected from his travels). He says,
I was unusually organised for a boys hostel room. But being a hardcore techie doesn’t mean you can't have a left brain and a social life.
The first job
Pramod gave his first interview for a job at Infosys in September 1993. Later, he gave two more rounds in October and November, going on to join the company at the end of the year. Back then, Infosys was around a couple of 100-people strong, operating out of Koramangala, Bengaluru. Pramod was training the new joinees at the company. The teaching experience helped him by strengthening his fundamentals even further. He recalls,
At that time, Infosys only recruited from the IITs. These recruits could really spin you around and give bad feedback.
At Infosys, in 1994 only, Pramod was programming on the internet (on CGI programming). That was the first time he came across Nandan Nilekani. Nandan figured out that Pramod was the only one programming on the internet at the time. This was the time before mosaic browser and when JAVA was barely starting to evolve. Nandan told Pramod that 'consumer banking will dramatically change with the internet', and asked him to build a prototype. Banks didn't even have websites back then, and even core-based banking wasn’t practised by them (brand-based banking was common). Pramod built the prototype, which demonstrated through mosaic browser how people could log in through a fictitious banking website and do their banking. Though the company didn't take the product to completion, the project gave Pramod a chance to work with Nandan for around three months.
In 1995, Infosys funded a startup in the Boston area called Yantra Corporation, a supply chain product. Infosys put in $100,000 seed capital and gained a 100 percent ownership. The company’s CEO, Devdutt Yellurkar (General Partner at CRV [Charles River Partners] now), was looking for a tech guy who had worked on the internet before. Nandan recommended Pramod. He moved to the US in 1996 to be a part of the company. He recalls,
It changed my life completely. Sitting in a small office in Massachusetts and suddenly walking into the office of the CIO of GAP, J.C. Penney, talking about how we're trying to build a product that will disrupt the supply chain network.
Joining Yantra meant giving up some potential stock options in Infosys. Pramod, however, took an instinct-based decision and took a plunge. He really enjoyed working with the people there. He was excited about working with a great CEO and mentor, and alongside a great team trying to do something very different.
Yantra Corporation raised two more rounds of investment, one in the late 1990s and another in the early 2000s. This was amidst the difficult times of the internet, with the dotcom burst. The company was very successful in retail. They rewrote multichannel retail commerce and owned the entire turf, including Hallmark, JC Penney, Big Best Buy, Target and GAP.
Pramod wrote a service-oriented architecture paper in 2003 and spoke on how micro-services and API-based platforms ease the way to build. In 2005, Yantra Corp. (then a $40 million company in terms of revenues) got acquired by Sterling Commerce ($600 million in revenues) for a 5x valuation. Pramod believes that it was primarily because of the superior product and engineering that the company got acquired at such a valuation. In a unique scenario, after the acquisition, the CTO, VP Engineering, Senior VP, CMO and head of product management at Yantra Corp. replaced those respective positions at Sterling Commerce. Yantra Corp employees remained loyal to Sterling and stuck around for a long time. Pramod stayed on as well, finally leaving in 2009.
Sterling Commerce later got acquired by IBM for $1.4 billion.
Identifying opportunity in identity
In 2009, when Dr Manmohan Singh invited Nandan to do an identity project, Pramod almost instinctively decided to get in touch with his old colleague at Infosys. He says, "I had read about the power of digital identity and why digital identity and digital systems will bring about transparency and reduce the friction created by human beings who created discrimination among individuals."
Nandan told Pramod that he was taking a risk, given that he was enjoying his work at the time and it was uncertain whether the identity project would even take off or pay him anything. Pramod replied that he knew intuitively the right thing to do, and shifted within a week to do full-time volunteering for Aadhaar in July 2009. Pramod believes that Nandan had this magnetic capability to bring the right people to work with him.
The basis of Aadhaar
When an opportunity is given to you, and you have the capability to do it, then not doing it is not acceptable in my mind. Trying and failing is okay. (With Aadhaar), we didn't know if we would succeed or not, but we believed that we could and that we'd do everything to get there.
The UID team started with project planning, tech design and talking to experts in the first few months, which was followed by the formation of formal committees.
Design of Aadhaar
Within a month, three things became clear about the design of Aadhaar:
- Fresh enrolment: The team had to start fresh (through enrolment) and not be dependent on any existing data (election commission).
- Biometrics: If it had to be unique identity, then it had to be biometrics-based.
- Keep it minimal: A digital identity and not an offline card-based identity. It had to be a general purpose identity platform, not a static/physical identity card (like the PAN card).
Discussion on the two aspects of identity - demographic data and biometric-based data - required a committee each. The demographic data committee was headed by N. Vittal, ex-Chief Vigilance Commissioner of India, while Dr B.K. Gairola (from NIC) headed the biometrics data committee. The demographic data committee finalised the four attributes to go into the database – name, date of birth, gender (including the third gender), and address. They were kept minimal to achieve inclusion.
Pramod’s key takeaway from his experience of working with the government is simple:
If you're genuine and if your intent is clear, it generally works out. This is the exact reason why I continue to work with UID today. You have to be truly appreciative of the way the system thinks.
Nandan also emphasises this – “If you want to make massive changes in the system, it doesn't come in a day, we cannot be impatient about it. We might also lose some battles, but we have to win the war.”
Tech stack of Aadhaar
Pramod recollects a rather worrisome conversation with a professor in the US who specialised in biometrics, having worked in the area for a long time (he volunteered for Aadhaar). He said,
You'll need a football field full of million dollar computers to do biometrics deduplication of more than a billion people. It is possible technically, but unviable economically (a million computers of a million dollars each).
Pramod then spoke to Google and Facebook and studied their architecture. He realised that in Aadhaar, one identity deduplication has to do nothing with another person’s ID deduplication of data. They realised that it's a very parallelisable problem. Pramod then took a personal bet and decided to go with an open source, commodity computing. They didn’t use any proprietary chipset or computers because they couldn’t have a (architectural) lock-in situation of national critical infrastructure like this (if it succeeds).
The team decided to go with the following:
- Open source
- Commodity computing architecture: Knowing the speed at which commodity computing was growing (doubling each year according to Moore's law), Pramod decided to bet on it.
- Unbundle: The team didn't want one vendor solution, so they unbundled the entire identity system piece by piece. This means unbundling the identity systems to enrolment, workflow, deduplication, authentication, error handling and manual quality checks.
In the short term (first two years), it looked like we were spending more money because we stuck with the open computing architecture. At the end of six years, the entire cost of the UIDAI system is Rs 70 per person, including hardware, data centre, salaries, offices, and every other cost involved. It was almost $100 in the US (per person) and 100 pounds in the UK.
Pramod keeps telling the following to the architects (design people):
- Understanding the purpose/problem statement. It comes above everything and everyone.
- Clarity in your thinking.
- Seeking feedback from many people.
- Keeping it simple (and stupid).
Techies and ground reality
There were situations where the tech team of Aadhaar didn't understand the ground realities. They didn’t know what happens in the village once the laptop goes and the camps (biometrics) get done. Though the original process of enrolment was simpler and technically correct, it didn’t take into account factors like the handling of machines (and devices) and poor connectivity. They eventually added failure resilience in the system, and after many iterative pilot-based run throughs and updating software design for enrolment, the biometrics data collection camps were organised.
The team also realised the need to include approvals in the backend workflow. They readjusted the manual process and built a manual verification and adjudication workflow.
Building India Stack
As chief architect of Aadhaar, and because of his exposure to the system and the ability to solve large-scale design issues, Pramod got access to build the rest of the layers of India Stack, the name collectively given to a bunch of layers the government has built. The first layer was Aadhaar and the second was eSign. This was followed by the submission of digital locker architecture paper, on top of which a bunch of payment layers were built. One of these is Aadhaar-based payment, through which the DBT (Direct benefits transfer) money is transferred. Looking at the uptake of IMPS in 2014, Pramod started working on an instant money transfer solution (layer), which is extremely simple and mobile-enabled to facilitate peer-to-peer transactions and merchants payments.
For this to happen, you need a powerful API at the back that allows a distributed but instant money transfer protocol. That, when I wrote the first UPI protocol, sort of did what SMTP did to e-mail.
However, email was different with respect to the following three characteristics:
- Store and forward,
- Doesn't have encryption between the end point and
- No trust establishment.
All the banks have access to UPI. And now with demonetisation, wallets will also join in.
After the UPI, Pramod is excited about the following forthcoming layers on Aadhaar:
- Electronic Consent Architecture: This will put an individual in control of their own digital footprints. It’ll provide a technological framework to provide a control (a consent) so that one can still use his/her footprint to get a loan or apply for a job. The draft for electronic consent architecture is out.
- GST: This will take the system from a portal-based thinking to API-based payments.
- Bharat Bill Pay Systems: Out of one billion utility bills paid per month, only 3-5 percent are paid electronically. There’s huge scope of easing the payments here.
- Electronic Toll Collection: This will enable using RFIDs in vehicles. It will be useful in situations like implementing a congestion pricing for vehicles travelling through a congested area through an API.
Desirable qualities in a tech person
While hiring a techie, Pramod looks for the following qualities:
- Depth of design thinking and why: Not the syntactical depth but the design thinking depth, especially for senior/architect level guys. Why we do something is more important than how we do it. How did the techie think about the design, network, storage problem?
- Passion for what you're doing: You've got to be alive and spread that energy to the team. If you want to be an architect who can influence people's thinking, then this is a must.
- Ability to communicate your thought process in simple words: Know something yourself doesn't mean anything. Getting people who are not technologists to understand why are you saying something is important.
People who influenced his life and value systems
In our conversation, Pramod kept on emphasising that he’s a people person. He is influenced by a set of people who are very close to him. Here’s what marks these people have left on him:
- Mother: She was an extremely passionate teacher of the Malayalam language. Pramod believes that his animated style of talking on stage (or while teaching) comes from his mother. He believes in talking from the heart and not from a power point, which he learnt from his mother.
- Father: He was a karmayogi, who focused on work, family, value systems, and kept them together. He had no rigid social structure or hierarchy in his mind for treating people differently. That lack of rigidity is very valuable to Pramod.
- Brother: Prajod has been his best friend throughout his life. He has been a huge support whenever Pramod was in need of it.
- Grandmother: At 99, Pramod’s grandmother is still going strong. She held the entire family together. Educated till the 4th grade (because princess couldn't study more), she used to teach Sanskrit to MA students later in her life. She authored books and taught the veena, all while running the whole house. She was also an amazing story teller.
- Wife: Pramod knew Lekha from the 11th grade, when she was in the 9th grade. She is holding their family together and has been a huge support. Since they knew each other from such an early age, they didn't have anything to hide or not talk about. This, in turn, allowed them to pursue many more confident ventures in life.
- Cousins: Pramod has some crazy cousins who are very close to him, thanks to his grandmother holding the large family together.
- Devdutt Yellurkar: He has played an instrumental role as a mentor, guiding him as a product architect.
- Nandan Nilekani: He’s a visionary from whom Pramod continues to learn a ton even now. He’s surprised by his ability to abstract a pattern from a scenario and say it'll work elsewhere. He believes that visionaries have this power of abstraction, which allows you to compare the two situations and use components of learning to solve bigger (and newer) problems.
If youngsters surround themselves with the right people and put their hearts into their work (rather than only the brain), very rarely will they fail. It'll help you discover yourself and do the right things.
You can follow Pramod on Twitter.
If you have recommendations for Techie Tuesdays, then you can nominate/suggest on this form.