Research in Industry

I am happy to have suggested Mr Sundar rajan GS to interview Dr. Sitaram Ramachandrula, Senior Research Scientist, HP Labs, Bangalore.

Here is an interesting interview with him on the topic "Research in the Industry".

Q: Please give a brief introduction about you.

SR: My name is Sitaram Ramachandrula, and I am a Senior Research Scientist in HP Labs India in Bangalore for last 5years. I have done my BE in ECE from Osmania university campus and M.Tech in Digital electronics from Cochin university campus and Ph.D in ECE from Indian institute of Science Bangalore with a specialization in speech recognition. Prior to working with HP Labs, I worked with Philips semiconductors and Philips consumer electronics divisions for nearly 5 years. Before that I had brief stints in Encore software and Ammanna technologies.

My research interests are broadly in pattern recognition and applications of digital signal processing. Specifically my areas of interest are speech recognition, document image processing, optical character recognition (both machine and hand printed) and location based services. I play table tennis when I get an opportunity.

Q: In the schools of the world, which do you suggest is the pioneering school that does the best and fierce research in your field of interest?

SR: As I mentioned, one of my area of interest is 'Speech recognition'. Carnegie Mellon University (CMU), MIT, AT&T Bell labs, IBM research labs, Cambridge University UK did pioneering work in this area. I will pick CMU for its contributions in this field. Lately, Microsoft research and Nuance are the largest speech research groups.

Q: For a new researcher - can u suggest shortly the steps that one should take in order to become successfully established in research. (What I mean is finding a topic, then a guide, then proceeding etc.)

SR: For students before embarking upon research, picking up an area for research: This is ideally done Top-down, but many end up doing Bottom-up based on opportunities available, I mean they apply to many top universities and based on the admissions they get, they choose the best in them according to some criterion. A Top-down approach where an individual really spends a considerable amount of time thinking, analyzing and doing an honest self-analysis e.g., what is worth doing? why? what he/she can do the best? where is his/her interest? This needs a considerable time (say 1 to 2 years). Many cannot do this well while doing their BE, so an alternative is do a job for 1 or 2 years after BE and get time to do this analysis. This needs to be done very actively, by reading many technical magazines, internet etc.

This process is needed mainly not to leave out topics which are more important to the society and industry. Of course the technology and knowledge are universal, but for few societies, few topics are more important than others, so those societies end up investing in those technologies the most. Now a days this sense of importance to society is getting neglected and we end up doing research based on outsourced job potential. For a new researcher, depending on whether he is doing an MS or a PhD, I first suggest them to get some guidance in choosing fundamental subjects/courses to deeply master in the first 6months to 1 year before embarking on the chosen research area.

The next step is to do literature survey, to establish/get a feel of the chronological developments in the area (over a span of past 20 years or in some cases more than that), to find out the importance given to different topics at different times in that area. And what are the current different topics which people are looking at. What are the popular approaches or methodologies of looking at the topics. This part according to me is a continuous exercise throughout the research, (even after choosing a topic/problem/approach), so that one is abreast of latest developments or even to get new ideas. Most importantly, people need to spend time thinking while each of the paper is being read as knowledge continuously dawns. This is needed, as the researcher needs to create a new path, which in his/her thinking is interesting or important and he/she ultimately needs to prove/validate his/her own path by implementing or by theoretically proving.

Researchers need to be ready for different types of work, reading as a student, reading as an expert, learn to implement, implement to learn, implement to test ideas, thinking, thinking big, thinking differently, facing failures patiently, spending months just brooding/thinking. All these phases will not be equally enjoyable or give immediate results but put together all train a person to be a good researcher, as any real problem, or any real solution is not going to be that trivial.

Q: How is research in Industry (research labs) different from that in universities? Is there any collaboration between the two?

SR: Industry research is more directed and done in larger groups compared to academic research groups or what a PhD or an MS scholar does. Industry can afford to do bigger and expensive experiments or take bigger risks. And many ideas can be pursued quickly in industry. More Ideas get generated dynamically, due to group brain storming. But finally when it comes to one individual taking up a research task, the pace will be similar as researcher is dealing with the knowledge and real problems. Only difference is in industry many take up those tasks and knowledge gets churned faster. Lastly industry research aims to produce business by creating value.

Collaborations are definitely possible, when there is a potential difference between any two groups, the knowledge flows in one direction and when both are similar, knowledge and idea exchange happens. Some problems are more theoretical or farther from business, and industry may not have time for them, until those problems reach a stage academic research can contribute to industry. And industry can sponsor that kind of efforts by seeing their long term potential.

Q: How does a student decide if he is more suited for "Academic Research" vs "Industry Research"?

SR: This is individual choice, industry research makes you do nearer to practical things, can make you richer economically. Academic research gives you freedom to do more profound research and that may not be nearer to market or so no market pressure but can make you richer in deep knowledge. A good researcher can anytime change to other type of research. Sometimes costly explorations attract academic researchers to take sabbaticals to industry, and more theoretical topics attract industry researchers to spend time in academic institutions/groups. Most of the academic researchers also teach, which in itself is interesting to some; apart from giving them opportunity to visit well known fundamental topics again and again, and also look at them w.r.t. current developments outside. This gives opportunity for bringing a new light into fundamentals. Lastly Industry research makes people to also continuously think for the other two often ignored directions of research by academicians, i.e., the business and usability of the technical research being done.

Q: Almost any research topic connected to engineering requires a student to be comfortable in programming - amongst the huge pool of programming languages (C, C++, FORTRAN, MATLAB, Mathematica, Perl etc.,) - what do u suggest is the most suited for research oriented programming?

SR: Ya I do agree, good programming or implementation skills are helpful in research. Depending on the topic of research the programming languages or tools become appropriate. MATLAB is good for many areas of Signal processing. But C or C++ and scripting languages like PERL etc are preferred in areas such as Speech recognition, here you have more control and get more knowledge of the field while developing algorithms using these languages. Based on the vastness of the area, the toolkits get gradually formed by the research communities, using these toolkits appropriate to each topic is a good approach. Of course these toolkits themselves get upgraded and a researcher is expected to learn to work on the source code of these tool kits.

Q: What do you think is the most vital and unique part of Research that can never be got in any other spectrum of professions?

SR: All the professions have their value in the society. Some professions are characterized by less range/dynamics of feelings they produce, some produce high range of feeling in the minds of people pursuing them. Some topics of research can take people to bliss, when on a right path, but researchers also need to be patient when failures occur which can take people to their lows (e.g., all your effort could turn out to be in a wrong direction, or your costly experiment failed or all of your great ideas have occurred to someone else few years back).

Historically society has faced many unique situations, and is going to face more and more unique and specific problems, the solutions to these are not going to be simple, each of these problems need good amount of research work. Unless some healthy proportion of population takes up (or is encouraged/enabled to) research (as a profession) in each field, we are doomed to solve any problem we face in the hard way, sometimes re-inventing or sometimes spending/loosing huge amounts of time and money. Lastly, research profession has potential to give higher value to the society and researcher can enjoy more respect and have more say on things ('Knowledge is power').

Q: In HP R&D, which field of engineering would find most applications like, amongst Electrical, Instrumentation, Chemical, Mech, Prod, Meta, Comp Sci - who would find better opportunities in the R&D in HP.

SR: In HP Labs India major chunk of researchers are from a) Electronics-and-Communications and b) Computer Science engineering.

Q: A student has decided to do research - how do u suggest that he does a self analysis to really find out whether he is interested or can really sustain himself by doing research. (Peer pressure is very common in colleges, so he might have resorted to research just because his friends are into it)

SR: As you said, a simple self analysis will indicate, whether he/she is capable or interested in a research career. A simple test of capability is: with the current knowledge/intelligence they possess are they able to follow at least one thread of the field till the state of the art, say in six months? This makes them ready to start thinking innovatively for problems being addressed. However, hard work is an important ingredient for a research career, which is easy to figure out in self. And results will surely come to motivate if the effort is good enough.

Profile:

Sitaram Ramachandrula is a Senior Research Scientist at HP Labs India working in PriDE project. He joined Hewlett Packard in March 2003. He has earlier worked in speech recognition for local languages. He has also worked in few projects in the areas of document image processing, video global motion estimation and location based services.

Prior to joining HP he was a senior technical leader with Philips Semiconductors, in Philips Innovations Campus, Bangalore, India, for over 4 years. Earlier he was an assistant project leader in Encore Software, Bangalore. He holds a PhD in Electrical Communication Engineering, with a specialization in "speech recognition", from Indian Institute of Science, Bangalore; M.Tech in Digital Electronics from Cochin University, Kochi, and B.E in Electronics and Communication Engineering from Osmania University, Hyderabad. His current research interests include: offline handwriting recognition and document image processing.