Coded, Created and Patented: Kaibalya Prasad and his Simplified Solution

Coded, Created and Patented: Kaibalya Prasad and his Simplified Solution

“I will do everything in the straight forward way, no shortcuts, logic or out of the box thinking”

                                                                                                                  -said no engineer ever.

Engineering on its own is the study of twisting things around, finding new ways, learning and innovating. As an institute of technical education, NIT Rourkela stands up to its name and fame. We boast of innovators, intellectuals, designers, scientists and several others who have excelled in the field of technology. This week Monday Morning brings you the story of Kaibalya Prasad Bhuyan, a final year student from the Department of Electronics and Communication Engineering. A former intern at Qualcomm, he developed an algorithm that simplifies the process of testing which earned him a huge financial reward. To know Kaibalya’s story in his own words, dig into the interview below. 

Monday Morning(MM): Walk us through your early life. How did you land at NIT Rourkela?


Kaibalya Prasad Bhuyan (KPB):  I hail from Keonjhar, Odisha. I completed my schooling from Saraswati Sishu Vidya Mandir and cleared my intermediate from a CHSE based college. After clearing JEE Mains, I was willing to join the Computer science branch at NIT Rourkela. But subsequently, I met some people who changed my perception towards it. I realized that topics that fascinate me the most, like signal processing, VLSI, IoT and all, were a part of EC rather than CS. So, it was an informed choice.

MM: Brief us about your patented algorithm


KPB: The algorithm is called dependency parsing. It was designed to provide a simple description of the grammatical relationships in a sentence that can easily be understood and effectively used by people without linguistic expertise who want to extract textual relations. In particular, rather than the phrase structure representations that have long dominated in the computational linguistics community, it represents all sentence relationships uniformly as typed dependency relations. 

MM: How did you get the idea for that? 


KPB: Previously we are working with Qualcomm parser to parse various parameters from a log file. But the searching or grouping algorithms used in their parser were not up to mark. Based on some queries the parser needs to crawl through the database and give some effective results. But the queries were like keywords, not a phrase or sentence type. If someone encountered a problem he needs to search what is the base of that problem, which may difficult for a common tester as he can’t remember the root of every problem. I took the idea from here, that is the parser needs some improvement so that a layman can interact with the database with some queries. More flexibility will be given to the queries i.e. you simply need to write the problem in English and then use that as a query. For a smooth experience, this diagnosis system can communicate through the internet or mailing system. 

 

MM: Brief us the procedure of your patent.

KPB: First I was offered for Qualcomm parent. A Qualcomm parent is a very big thing but one demerit of this outstanding program was, one will lose complete ownership of that algorithm. One of my colleagues suggested me to file a patent on your own. And the greatest advantage was he had been in a parent company 1 year ago. He helped me a lot during this complete process. Some third-party companies can take care of everything and contact directly to the U.S. Patent and Trademark Office (USPTO). They filled my parent application. Then I wrote a paper about the invention.

You cannot get a patent just based on an idea. You must show how your invention works. In addition, your invention must be new. This means it must be different in some important way from all previous inventions in that field.

The thing that made this invention better was long term dependency. And the thing is Applying for a patent is a complete business decision. Then all the documents comments of reviewers were submitted to USPTO and then after some days they granted. 

MM: How did the private company help you in patenting the DNN code? 
KPB:

The major advantage of this algorithm is deep text searching. For beginners, text analysis often begins with a simple keyword search. For advanced text searching, however, you need to have information about word relationships. Here deep parsing will come into the picture. Along with that information extraction, comes text summarization, sentiment analysis and of course argument mining.

Some database based companies need these algorithms to get an efficient result. This algorithm is a breakthrough in the language modelling field. The best example is the next word prediction in the android keyboard. This model beats some great existing language models in terms of long term dependency. 

MM: Why didn't you patent your code with Qualcomm?


KPB: This was a personal choice. I want to go for Microsoft that’s why some of my colleagues suggest that if you get it patented on your own then it has more value than Qualcomm patent. 

MM: How beneficial was your Qualcomm on-campus internship? 


KPB:  Frankly the internship at Qualcomm is my first corporate internship and I got a great experience. The testing profile didn’t suit me initially but their experience, experiment setups, measurement devices, mind-boggling algorithms fascinated me very much. Every day I went to lab feeling this could be one of my life’s dreams. But after some days I was going through some set of experiments.

MM: What was your field of work at Qualcomm? 


KPB:  Initially I was assigned to the testing team. My life as a tester at Qualcomm was rather interesting. At the beginning of the day, I used to get an mail from the team about my to-do list. Then I needed to do those testing things within time. Specifically, my work was LTE call flow testing. There I was observing some parameters during the LTE call flow for every possible combination. There were a bunch of codes that were already written and some software for diagnosis and process control. I was assigned to run those files in some particular sequence to get the results, if I were getting some technical error then I was told to report them to the debug team. In one line I can say that you don’t need to put your mind anywhere. Like literally turn your brain off.

But I secretly did some crazy stuff. Like as there is no point of thinking what was the process why was this process is going, so I wrote an AI problem which can read my Outlook emails and prepare the to-do list. And then execute them according. This helped me during the mid evaluation. After seeing my work during the evaluation some people in the panel asked me what I was doing in the testing department and rather I could be helpful in the development team. Then I got into the research field. And there all this happened to me.

 

MM: What inspired you to choose a career in the research field?


KPB: I will give the full credit to the Qualcomm research team. I started my research career with deep language processing. It involved automation of the testing things by an email, how to debug the failed test cases, how to write python code from the general English language etc. Initially, Qualcomm employees took a step into this but they failed. After my midterm work, they thought let’s try that again. I wanted to automate the whole testing thing as it is so boring for people like me and the vision was very tough and challenging. These things were the main motivation and of course very inspiring for me at that moment.

MM: Brief us about your various departmental projects. 


KPB:  For the idea, I am very much thankful to the AIIMS project manager. I remember a line he told us –

As a video processing engineer, you can do everything if it were captured in a frame.

From there onwards I involved with some departments as they captured their experiments into a video file. Starting from vehicle tracking to generic mutation study I have been a part of major video processing research. The major civil department project involves lane detector, camera zoom in zoom out, perspective transformation, super-resolution, image reconstruction, unmanned vehicle, etc. The major genetic mutation department project involves locomotion behaviour study, statistical analysis of behaviour, microparticle tracking, etc. In my departmental project, I took brain segmentation from MRI imaging.

MM: What are the other projects you a part of? 


KPB:  AI is becoming a critical part of modern welfare. Compared with conventional systems, military systems equipped with AI are capable of handling larger volumes of data more efficiently. Additionally, AI improves self-control, self-regulation, and self-actuation of combat systems due to its inherent computing and decision-making capabilities. Almost all top country is now trying to mix AI with their military system, recently Microsoft won the contract for weapon and security modification of the US defence system. So why not India. As a part of the DRDO AI team, I am also involved in this party. We are modifying various existing range technology to make India's defence system strong. The projects I am being part of are auto-lock weapons with full functional GPS and camera tracking system, AI-IOT based decision maker weapon, Target Recognition, Combat Simulation & Training, etc.

MM: Brief us about your recent project under the MHRD.


KPB: Currently I am working with feature control from a satellite image. This project is sponsored by ISRO. Using deep image processing we are segmenting the image based upon some queries. Satellite images are very high-resolution image and hence difficult to process also there are many obstacles which makes it more challenging. 

MM: What all fields are you interested in and what is the ultimate field you want to get into?


KPB:  Ever since my childhood, the wretched health care system in our locality made me wonder if I could somehow contribute to making the diagnosis and treatment of various diseases easier by the aid of technology. After learning about the concepts of AI, ML, DNN, etc. I came up with the idea of integration of these technologies with medical imaging systems, which will help in making the diagnosis of many dreaded brain lesions easy, accurate and rapid, which will aid the health care professionals to save a lot of lives.

The ultimate goal in my professional life is to contribute to the field of emotion translation, which piques my interest.

MM: What are the other various fields you devoted your time to? 


KPB: Coming home after work I like using my free time to sit around playing video games and watching TV series. Along with that I love travelling and hiking. Expressing my feelings and thoughts through a pen is one of my favourite things to do. 

MM: What are your future prospects? 


KPB: I have a lot of dreams. This question is similar to what are the prospects and application of deep learning. I want to be a major part of this development. The future of anything is uncertain, you can’t predict precisely. But as per the more innovative developments in the same field, you can envisage the trend and forecast various opportunities lying in the related fields. Making autonomous vehicle successful, smarter virtual assistant and chatbot, shifting towards unsupervised based learning process, and all are my current dreams. But the things are dream can’t be constant it will change if you will meet some breakthrough concepts, maybe some people or something like that. My ultimate goal is to create an AI hospital.

MM: What message would you like to give to our readers? 
KPB:
 

Never be disappointed by any setbacks you face in life. Just keep moving forward with patience and perseverance and continue doing whatever you love. One day, you will undoubtedly get a breakthrough and when you do, never be afraid to do something that will benefit others.

 

The most fascinating aspect of life is that it has no boundaries, it can evolve and adapt as per its circumstances. In a constantly developing and changing world, innovators like Kaibalya Prasad Bhuyan break the chain of monotony. Monday Morning wishes the best to Kaibalya Prasad for all his ventures ahead.  

DISCLAIMER: The content, opinions or views expressed on the Monday Morning's website and its social media platforms, including, but not limited to Facebook, Instagram and Twitter pages, are strictly the property of Monday Morning and represent the extensive research and work of the working team of respective academic year of Monday Morning and not those of the institute. The reports and statements published are consolidated from the collected background research and interviews. The institute's official statements can be found in the press releases published by the institute or via an RTI application.

No article or any statements by Monday Morning is to be reproduced, presented or distributed in part or whole without prior permission of the Executive Body of Monday Morning for any purposes, including, but not limited to print and electronic form.



Comments

    Leave a comment

    Login to comment.
    Ask a Question Forum