Indian IT Industry

[Supratik]

India's largest Edutech compnay Byju's

https://www.linkedin.com/pulse/byjus-in ... mana-mitra

[uskumar]

There are probably hundreds of PCB factories in China, not sure about India but I would be very surprised if we don't have at least a few.

Till few years back there were almost no PCB factories in India. But since introduction of tax on imported PCBs there has been steady uptick of the number of PCB factories.
import duty imposed on populated PCBs in April
Due to which

xiamo sets up PCB factories

[uskumar]

nice to see RBI being firm in Data localization.
Now to the next step. Localization of Personal data including emails.

Things government can do to complement rbi data storage

[souravB]

nice to see RBI being firm in Data localization.
Now to the next step. Localization of Personal data including emails.

Things government can do to complement rbi data storage

RBI and in extension GoI for other verticals is not doing it out of the goodness of their heart or even from data security point. If you read the article fully it very clearly says that GoI is happy with just a copy of the data stored here so that they can easily look into it whenever they want.
The points that have been raised in the article regarding what stopping government to just access the data wherever it is and not have a copy here by law doesn't serve any purpose for GoI. They would have to go through legal channels to access data rather than pick up whatever and whenever they want.

Mods: This may feel like I'm ranting or trolling. If you deem so please delete or remove this post.

[isubodh]

RBI and in extension GoI for other verticals is not doing it out of the goodness of their heart or even from data security point.

Agreed, but these kind of non-Tariff barriers have been effected in EU, China and US, keeping our exports down. I believe it will bring in few hundred jobs more to India then its good. It also aid in more cloud providers setting up shop and datacenters here. Even using it as a point of negotiation in service exports will be good use.
This is effectively what we are trying to do in Electronic Hardware too. If we have a market companies will come. We can care about Data Security later when we have established a sizable market, but reverse will never work out.

[arshyam]

The data protection draft asks for the "operational copy" to be stored in India. That's a clever way of going about this...

[hanumadu]

Is Machine Learning still considered hype or is it here to stay?

[ArjunPandit]

Is Machine Learning still considered hype or is it here to stay?

It's here to stay, but it will evolve. It's basically commoditizataion of a luxury called predictive/statistical/time series modeling.

[VinodTK]

Intel to open technology centre in Hyderabad

Hyderabad: Telangana continues to attract investments into the State. The latest to add to the list is the marquee technology giant Intel which will soon set up its technology development centre in Hyderabad.

A top delegation headed by Nivruti Rai, country head, Intel India, met Telangana IT Minister K T Rama Rao at Camp office in Hyderabad on Friday.

In the meeting, Intel delegation and the Minister talked about the industrial policy of Telangana and the infrastructure of Hyderabad, which encouraged the tech giant to choose Hyderabad to set up their tech centre.

“We are very happy that Intel decided to start a tech center in Hyderabad,” said KT Rama Rao.The Intel India centre in Hyderabad will provide a major boost to the IT sector in the region. The proposed centre is expected to create 1,500 jobs in its initial stages.

The government of Telangana and Intel delegation had held several meetings which culminated into setting up of the centre in Hyderabad.

The delegation invited Minister K T Rama Rao to attend Intel India’s 20th anniversary scheduled on November 15 in Bengaluru.

[Singha]

Is Machine Learning still considered hype or is it here to stay?

It's here to stay, but it will evolve. It's basically commoditizataion of a luxury called predictive/statistical/time series modeling.

the top data scientists now enjoying the fattest pkgs in india are mostly from maths/stats backgrounds and entered into ITvity after prior careers in places like indian statistical instt, india meterological bureau monsoon data crunching, Actuaries in insurance & loans sector building risk and mortality models. they were already deeply grounded in the maths and stats theoritical knowledge and knew most of the data analysis and mining techniques @ scale.

ofcourse not so cool sounding as ecommerce or ads based free model of data mining on people.

the online courses now claiming to teach data science are just going over the bare basics of the underlying theory and focussed on tools and libraries probably.

[SBajwa]

It's here to stay, but it will evolve. It's basically commoditizataion of a luxury called predictive/statistical/time series modeling.

the top data scientists now enjoying the fattest pkgs in india are mostly from maths/stats backgrounds and entered into ITvity after prior careers in places like indian statistical instt, india meterological bureau monsoon data crunching, Actuaries in insurance & loans sector building risk and mortality models. they were already deeply grounded in the maths and stats theoritical knowledge and knew most of the data analysis and mining techniques @ scale.

ofcourse not so cool sounding as ecommerce or ads based free model of data mining on people.

the online courses now claiming to teach data science are just going over the bare basics of the underlying theory and focussed on tools and libraries probably.

With lots of "Junk" data aka also called "Big Data" being collected everywhere (think of cars, phones, fridges, televisions, game consoles, watches, health meters, hospital equipment, planes, washers, dryers, etc) companies have no clue what to do with this data and thus dire need for Data Scientists. so get a degree in math and be a data scientist. Boeing improved the performance of their engines through data analysis saving 25%.

[Singha]

even the 24x7 health monitor bracelets w/A-GPS is mining and uploading a fair amt of data wrt to our health and whereabouts.
apple watch will take EKG. soon probably sonograms and other stuff. a woman who has conceived and does not know of it yet will get ads for baby products. a balding man for hair oil, a aeging man for ****.

we are being watched and analyzed in minute detail.

[ArjunPandit]

It's here to stay, but it will evolve. It's basically commoditizataion of a luxury called predictive/statistical/time series modeling.

the top data scientists now enjoying the fattest pkgs in india are mostly from maths/stats backgrounds and entered into ITvity after prior careers in places like indian statistical instt, india meterological bureau monsoon data crunching, Actuaries in insurance & loans sector building risk and mortality models. they were already deeply grounded in the maths and stats theoritical knowledge and knew most of the data analysis and mining techniques @ scale.

ofcourse not so cool sounding as ecommerce or ads based free model of data mining on people.

the online courses now claiming to teach data science are just going over the bare basics of the underlying theory and focussed on tools and libraries probably.

fattest packages as compared to IT industry. Start up, while being the largest, has its own payment dynamics so wont go in there too much. You should have heard the packages or billing rates that were being charged between 2000-2007. Today's rates are nothing compared to that.
What I have seen based on my experience in this field, big data binge will come to bite the industry big time. There's a reason FS can't use it and is not using it that extensively for core operations. . It will be used extensively for marketing, fraud risk prevention, to some extent for credit. I expect it to permeate at least for simulation based market risk models. For a lot of dynamics is not well understood and leaves the station by the time risk managers or quants will get a whiff of if let alone incorporate in a structured feedbak.
Yes there is lot of junk data, as Taleb said. it is like searching a needle in haystack. Big data has made the hay stack much bigger but needle still remains one.

[Vips]

How India is contributing in development of new-age computers.

Casual visitors to R Vijayaraghavan’s lab can be easily dazzled by his fancy equipment, one of which can cool bits of matter to temperatures well below that of deep space. His prize exhibit though is a tiny device that he has built, a proto-chip if you will, which he is developing to be part of powerful quantum computers in the future. Such a computer does not exist at the moment. Whoever builds one early will have a tool to dominate the world.

The foundation of quantum computing is a unit called the qubit, the basic unit of information in a quantum computer. Making qubits and getting them talk to each other is a hard task, because they are touchy and easily give up their quantum-ness when in contact with the outside world. Connecting them and making them work together is a harder task. Vijayaraghavan, a professor at the Tata Institute of Fundamental Research (TIFR) in Mumbai, has made a three-qubit processor where each qubit is connecected o each of the other two qubits.

In terms of the number of qubits, it is trivial. Early this year, Google announced that it had a quantum processor with 72 qubits, and IBM had demonstrated a 50-qubit processor last year. But their processor architectures are different from that of the TIFR device. In the IBM-Google approach, a qubit is connected only to the neighbouring qubits. In the TIFR device, every qubit will be connected to every other qubit. If Vijayaraghavan manages to scale his is processor to more qubits, he may have a processor that is very efficient in translating algorithms.

Quantum computers are entirely different from the so-called classical machines, as they are capable of solving problems that are impossible for today’s computers. In this sense, they are not super versions of today’s computers. A supercomputer, no matter how fast, cannot break security codes used in military and business communications. A quantum computer can break them in minutes. If, for example, someone in the world develops a quantum computer, all of today’s security codes become worthless. We would then need to design new ones using quantum computers.

In the global race to build quantum computers, India has so far been present only in theory compared to US, China and the handful of other European countries that were spending large amounts of money. India had no national programme. It had a number of theorists, but only a few had been trying to build a quantum computing device.

The basic need of a computer is a property with two states, like the presence or absence of current in a transistor in contemporary computers. Vijayaraghavan uses the presence or absence of an oscillation in a device kept at low temperature. Some physicists have tried to use spin, the way a particle aligns itself to a magnetic field, as a qubit. At the Indian Institute of Science Education and Research (IISER) in Mohali, Kavita Dorai is using spin as a quantum state, and has run quantum algorithm on a five-qubit device.

At IISER Pune, Umakant Rapol researches how to network quantum processors while TS Mahesh looks at quantum information and artificial intelligence for quantum control. The Indian Institute of Science (IISc) in Bangalore has several groups that work together. For example, assistant professor Vibhor Singh is researching on ways to build quantum processors in two dimensions, instead of 3D, as a better way of scaling them.

Now, promise of increased funding gives Indian researchers an opportunity to catch up with the rest of the world. “We are late,” says Vijayaraghavan, “but not so late that we cannot make a global impact.”

The power of quantum computing comes from the seemingly bizarre nature of matter at the atomic level. Classical computers store information in two states, either 1 or 0. In the classical world, which we all see and touch, one thing can exist in only one state at a time. In the quantum world, a particle can exist in two states. Not either 1 or 0 but both. Or other states in between. All at the same time.

Due to this property called superposition, quantum computers have the ability to do an extraordinary amount of calculations simultaneously. It can factor large numbers quickly, a feature that is important for information security. Current security systems use factors of large prime numbers as codes, but quantum computers can decode them easily. When codes can be broken easily, we need to use quantum mechanics to make other codes that cannot be broken by quantum computers. Which is why it is important to develop this technology for business and national security.

Quantum computers will also let us simulate nature at unprecedented levels. What happens really at the molecular level inside the body? Quantum computing will help simulate molecular interactions so well that we can design super-drugs that can cure diseases with little or no side-effects. “These are technologies no one is going to give you,” says Umakant Rapol, associate professor at IISER in Pune.

Among other things, Rapol is developing methods to scale quantum computers to useful levels. So far, one of the biggest problems in quantum computing is our inability to put together a large number of qubits to form a working computing device. For some reason, the qubits do seem to lose their quantum behaviour when they are put together.

None of this research is going to result in a commercial quantum computer in the country soon. But developing the technology can have other uses, many of which are strategic. For example, quantum computers are supposed to let us develop extremely sensitive interferometers that can provide precise locations of objects, which, in turn, can be useful to anything from leaking pipes to forecasting volcanic eruptions.

In the end, a large number of strategic industries are going to be based on quantum computing and related sciences. “In the 21st century,” says Kavita Dorai, “there is a clear advantage for nations having information. To use this information, we have to invest in quantum information and related technologies.”

India is now taking small steps towards laying the foundations of this technology. “Money may induce those who are working in the periphery to move into this area,” says Ashutosh Sharma.

[hanumadu]

https://timesofindia.indiatimes.com/bus ... 662725.cms

India would do well to spend more on its higher education and increase the quality of education. Both govt and private institutions should focus on quality by improving faculty, infrastructure and research.

[Vips]

Atos wins deal to build supercomputers for India.

French IT services firm Atos has won a three-year contract to build the first phase of supercomputers under India’s Rs 4,500-crore National Supercomputing Mission (NSM) as the country looks to create a cluster of machines for weather forecasting, drug discovery and data mining.

The tender to build these highperformance computers (HPC) had been floated by the Centre for Development of Advanced Computing (C-DAC) in February. Atos would be deploying its energy efficient Direct Liquid Cooled BullSequana supercomputers in India.

The contract is for all three phases of the NSM which would be running simultaneously, said C-DAC director general Hemant Darbari. “In Phase I, we will be assembling three supercomputers at an outlay of Rs 68 crore. In Phase II, it will be an aggregate of 10 petaflop, but the number of computers is yet to be decided,” he said.

The NSM is divided into two key tracks, build and buy, which are being spearheaded by the C-DAC and Bengaluru’s Indian Institute of Science respectively. Atos has won the contract for the ‘build’ part of the NSM for which it will partner CDAC in all three phases of the project. While Phase I involves assembling of the supercomputers, in Phase II, certain components like the motherboards would be manufactured locally, and in the third phase, the supercomputer would be designed in India by C-DAC.

In the first phase, IIT-Kharagpur will have a 1.3 petaflop machine and IISER Pune and IIT-BHU will have a 650 teraflop computer each. Atos will source the components from France and assemble the supercomputers in Chennai. “Work will happen simultaneously on all three phases, but each has different stages of delivery. The deliveries under Phase I will start in the first quarter of next year while

Phase II would be around Q2 or Q3,” said Arvind Bajaj, head of Atos Bull India.

When exactly the first set of supercomputers are delivered will also depend on whether the data centres at the institutes are ready. While C-DAC has already put out a tender for these, the contract has yet to be awarded.

Atos has a sizeable presence globally in the High performance computing (HPC) business under the Bull brand. In September it unveiled its most powerful supercomputer with a compute power of 12 petaflops in Germany, making it the country’s most powerful supercomputer, coming in at number 16 on the recently released Top 500 supercomputer rankings. The company has a supercomputer and server design centre in Bengaluru, which will work closely with C-DAC on subsequent phases of the project.

The NSM is being jointly implemented by the Department of Science and Technology and Department of Electronics and Information Technology at an estimated cost of Rs 4,500 crore over a period of seven years. The mission was announced three years ago with an intent to propel India higher in the global supercomputer rankings. The aim was to build a grid of 73 supercomputers across academic and research institutes in India, working on a wide range of applications, from weather forecasting to disaster management. HPC is important in order for India to make significant progress in various areas, from drug discovery to astrophysics and bioinformatics.

CDAC built the Param series of Super computers more the 2 decades back. Why is it again buying from abroad?

[chola]

CDAC built the Param series of Super computers more the 2 decades back. Why is it again buying from abroad?

It was always about designing the architecture. The components, especially the chips, were always sourced from abroad. CDAC is not a foundry or switch manfacturer.

[Vips]

Even now it is assembling, designing and just making of motherboards. This after 2+ decades...

[Prasad]

You expect CDAC to setup a billion $ foundry to make a few thousand CPUs and RAM?

[Singha]

IISC proposal for a 3500 cr foundry for research and strategic uses is the real deal.
it is still awaiting GOI approval.
and GOI still has not spent money on upgrading the chandigarh fab to somewhere near world class nanometer tech. our strategic products could benefit from newer techs instead of just sturdy old stuff.

[Vips]

You expect CDAC to setup a billion $ foundry to make a few thousand CPUs and RAM?

Yes, If it can help in ensuring security of our strategic systems, what is a billion for a $2.8 Trillion economy? Semi Conductor complex can be beefed up and its operations can be expanded.

[Prasad]

Please. Entire NSM has a capital outlay of $650m. IISc is still waiting for its $400m for a new GaN foundry. But yeah sure, goi is going to entirely fund $1bn for a brand new foundry. Despite dumping two earlier foundry building attempts.

[Bart S]

ATOS owns Bull and might have some hardware expertise but it is just a services company like TCS at the end of the day (perhaps one with more strategic technology smarts and not just cost arbitration driven like desi ones), this purchase has nothing that CDAC and a consortium of Indian companies could not build themselves with imported components. Mostly off-the shelf components are used in the majority of HPC systems these days with the only customized development being in the areas of software and interconnects, both well within our reach - it is not like we need to build the processors for them. I think they just took the easy way out in this case, and we didn't have the luxury of doing so earlier, which is why CDAC did some actual work in the 80s and 90s.

[Vips]

More Info

India has begun developing applications for cancer research, genome sequencing and drug discovery to be run on high-performance computers that it has contracted to French IT services firm Atos to build as part of the National Supercomputing Mission.

The applications include genome sequencing, simulations for cancer research and protein folding and breast cancer research in collaboration with Tata Memorial Hospital and the All India Institute of Medical Sciences, Hemant Darbari, director general at the Centre for Development of Advanced Computing (C-DAC), said.

French IT services company Atos won the contract to build the high-performance computers, the first of which it says will be ready in the first quarter of 2019.

The Rs 4,500-crore NSM envisions a network of over 70 connected supercomputers at academic and research institutions across the country, which will aid India in areas ranging from weather forecasting to drug discovery and astrophysics, among others.

India has 32 supercomputers across institutions with a combined capacity of 12.77 petaflops with two machines—Pratyush at the Indian Institute of Tropical Meteorology in Pune, and Mihir at the National Centre for Medium Range Weather Forecasting in Noida—ranking in the top 100 supercomputers in the world. “We have been working on various applications, which are a key part of the mission, as well as a few research projects including a high-speed network, Trinetra, that would connect the supercomputers once they are up,” Darbari said.

It is currently developing applications internally and testing these on the Param Shavak computers such that they would be ready to be used as soon as the supercomputers are ready. Param Shavak are the smaller ‘supercomputers in a box’ that C-DAC provides to academic institutes across India.

C-DAC is testing the network on two parallel bands of 240 and 600 gigabits/second, six to 15 times higher speeds than the existing network. The agency is also working on 64 core high-performance computing (HPC) processers, which would be used in the subsequent supercomputers as part of the ‘Make in India’ component.

[sooraj]

^ Is it AMD Zen2 64 core processers ???

[Karan M]

IISC proposal for a 3500 cr foundry for research and strategic uses is the real deal.
it is still awaiting GOI approval.
and GOI still has not spent money on upgrading the chandigarh fab to somewhere near world class nanometer tech. our strategic products could benefit from newer techs instead of just sturdy old stuff.

Its not that bleak. A lot of stuff was cleared and is in production already, we'll see the results within the next few years itself.

[nam]

A GaN foundry has been set up at GAETEC by DRDO. Ofcourse these may not have the miniaturization of nano scale, however they are enough for our defence needs. We already have GaN L, S, C band TRM and definitely X band SSPA.

Another aspect is these are power electronics devices. The research on GaN in IISc was funded by DRDO, and I believe the same team is requesting for a foundary at IISc.

The difference i believe is that the one at IISc is planned as a general GaN foundary, rather than focusing on power electronics like GAETEC.

[isubodh]

The Rs 4,500-crore NSM envisions a network of over 70 connected supercomputers at academic and research institutions across the country, which will aid India in areas ranging from weather forecasting to drug discovery and astrophysics, among others.

Can this type of power be created using GPU's/TFU's on say Google cloud ? Will having dedicated machines be cost-effective than hiring on cloud ?

[chaitanya]

^ You could use commercial cloud facilities for many academic purposes - I know AWS is used a lot in genomics. However, cloud systems may not be useful or optimized for all applications, especially for massive simulations involving climate modeling, etc. For example, I am not sure how useful a google TFU would be in CFD applications. Lastly (and most importantly), you may not want to run everything on a commercial cloud for national security concerns.You don't want AMCA or hypersonic engine design simulations to be run on a commercial service!

[Prasad]

The Rs 4,500-crore NSM envisions a network of over 70 connected supercomputers at academic and research institutions across the country, which will aid India in areas ranging from weather forecasting to drug discovery and astrophysics, among others.

Can this type of power be created using GPU's/TFU's on say Google cloud ? Will having dedicated machines be cost-effective than hiring on cloud ?

You want nuclear design related stuff run on google cloud?

Besides, one of the primary aims of the NSM is to improve our own capablity (even if we're assembling them with foreign-made products). Can't get that with hiring google cloud or aws.

[Singha]

imo Govt can create its own academic/strategic clouds and give time on that to all indian institutes instead of penny packet centers in all places, have some big meaty ones in each location and let people create their own partitions and timeshare.

this will be efficient vs attempting to kit up every institute to AWS level.

[SBajwa]

even the 24x7 health monitor bracelets w/A-GPS is mining and uploading a fair amt of data wrt to our health and whereabouts.
apple watch will take EKG. soon probably sonograms and other stuff. a woman who has conceived and does not know of it yet will get ads for baby products. a balding man for hair oil, a aeging man for ****.

we are being watched and analyzed in minute detail.

We are just scratching the surface. A university contacted me long time ago that they have collected cancer data from around the world from last 200 years in various formats (microfilms, paper, audio, etc) and wanted to digitize it converting into Data Warehouse (just for cancer)

[Neshant]

The Rs 4,500-crore NSM envisions a network of over 70 connected supercomputers at academic and research institutions across the country, which will aid India in areas ranging from weather forecasting to drug discovery and astrophysics, among others.

Can this type of power be created using GPU's/TFU's on say Google cloud ? Will having dedicated machines be cost-effective than hiring on cloud ?

How about paying Indians to use their cell phones, laptops and computers for processing time.

When phone is not in use, a screen saver runs with problems to solve sent from a server and solutions sent back.

SETI uses this kind of distributed computing along with many other such groups in the US.

That kind of processing power would dwarf any cluster of supercomputers.

[uskumar]

Data localisation clause may hurt trade pacts between European Union and India, says Ralf Sauer
This is first explicit threat to it/outsourcing from US or EU due to localisation policy, hence posting this

[Singha]

why exactly does it hurt US/EU cos to keep a small datacenter here and a few people with clearances to manage this data on indian consumers?
they could even rent the space within cloud providers here. its not really a big deal.

we should tell the EU/xyz to stfu and go with the same threat to US and China to whose demands they had no voice.

[uskumar]

It is interesting that the threat is coming from EU not US. When was last time an EU app was popular in India. How is this effecting them. Seems like they are doing dirty work of US.

[Rishirishi]

why exactly does it hurt US/EU cos to keep a small datacenter here and a few people with clearances to manage this data on indian consumers?
they could even rent the space within cloud providers here. its not really a big deal.

we should tell the EU/xyz to stfu and go with the same threat to US and China to whose demands they had no voice.

The companies usually have a single system with all the data. It is usually difficult to split the data, based on geography. The alternative is to replicate the data in all countries the company has presence. This is not as easy at it seems if you have 15-20 worldwide locations.

[Singha]

Its not that difficult too given the vast database expertise of the popular apps and faangm titans

there is a std tech called sharding where parts of databases are kept across many nodes. and what goes where can be filtered on many criteria incl perhaps geographic location inside the record. dropbox and anyone else with large databases that can service file download requests from clients use such techniques to keep it optimized i believe.

https://docs.microsoft.com/en-us/azure/ ... s/sharding

https://docs.mongodb.com/manual/tutoria ... -location/

the hypocrite cos who are dragging their feet wrt india have already GUBOed long ago for china and EU

[Vips]

Infosys opens hub in Hartford, hires 7000 Americans so far.

[Neshant]

Infosys opens hub in Hartford, hires 7000 Americans so far.

But if all they are doing is outsourcing rather than research & development...

OTOH if they are doing R&D, what are they researching and developing?