[Portrait Alumni PhD] Imran Sheikh, AI researcher à Vivoka

 

I graduated with a degree in Electronics Engineering in 2007 and started working for a multinational technology company specializing in information technology services. Luckily, I was chosen to work with an R&D team that specialized on the topics of my interest, which are Natural-language and Speech-signal processing. In simple terms, these are subfields of computer science that have enabled modern AI assistants like Siri, Alexa, Google Assistant, etc. I was a research engineer working on different kinds of problems like voice interfaces, speech biometrics, conversation analytics, low-resource speech recognition, and some others. While everything was going great, a few years down the line I repeatedly felt the need to expand my thought process. I decided to do a PhD and I chose to do it at the Université de Lorraine as it has this lab named LORIA which excelled in the areas of my interest. I did my PhD during 2014-2017, on certain problems encountered in automatic speech recognition. After my PhD I worked as a research scientist in the industry for 2 years, leading research on spoken language understanding, building IP for the company as well as consulting on projects involving voice assistants. I then switched to an exciting EU funded project which aimed at building technology for cost-effective, multilingual, privacy-driven voice-enabled services. At the end of this project, I took up my current position of an AI researcher.

 

 

I have had a remarkable experience at the Université de Lorraine. I had the opportunity to learn from some of the most intelligent and knowledgeable people, supported with one of the best-in-class academic facilities and infrastructure. At the same time, the university campus introduced me to so many new cultures, traditions and ideologies, through its students, researchers and associates coming from several different origins and professional backgrounds. All of these have played a vital role in helping me to rebuild my professional competencies, sharpen my research skills as well as to broaden my mind.

 

 

I am an AI researcher at a French voice technology company named Vivoka. Vivoka offers an all-in-one development solution that enables any company to create its own high-performance, secure and offline voice AI assistant that can run on embedded & edge devices. Technologies like speech recognition, speech synthesis, voice biometrics, audio enhancement and language understanding can be very easily integrated into existing applications and deployed on devices like smart glasses, helmets, robots, handhelds, etc. Unlike the cloud-based voice AI assistants available in smart-phones and smart-speakers, Vivoka's voice technologies run on-device and offline, and are targeted for use-cases like field services & maintenance, logistics & supply chain, healthcare & medical.

 

At Vivoka, I lead the research on speech technologies. My overall goal is to keep improving the performance of our embedded speech recognition technology. While the goal is quite straightforward, it involves several interesting challenges which keep me excited to go back to work each day. It involves different kinds of tasks, for instance keeping a check on the recent trends in both academia and industry, building IP for the company as well as new technology with USP, helping address core technology issues and feature-requests from clients. Moreover, there is always the opportunity to collaborate with academia, supervise interns and students, as well as to publish and review research articles and publications.

 

 

I have been working in this field for more than 17 years now and I think my interest in technologies for human-machine interaction, and specifically vocal interaction, has kept me deeply attached to it. Moreover, this field evolves quite rapidly and the transition from research to technology-in-use is fast, sometimes only a few months. I think striving to build the next version of a cutting-edge technology drives me further to work in this sector.

 

 

There are several qualities that are necessary, including the most typical ones like analytical ability, creativity, agility, goal orientation, and so on. However, I think one of the most essential qualities that is needed is open mindedness. As a researcher we are experts in our technical area but this expertise alone is not sufficient in building a successful and usable technology or product. The process rather involves inputs and feedback from several stakeholders and participants, including non-technical members. Open mindedness helps us to look outside our relatively small technical world. It improves our ability to communicate and collaborate, and to build something which is guaranteed to be even better.

 

 

In the past I have contributed with state-of-the-art deep neural network based speech recognition models for embedded devices as well as decoders, which are super-efficient software programs that carry out the decoding of speech signals into text in faster than real-time on small and low power edge devices. These were built right from scratch, along with a super-brilliant colleague (Vincent) at Vivoka. In the last few months we have shifted our focus to something different and innovative. I will be able to talk about it after we have officially made it public.

 

 

There are several things that I appreciate about my job. I think the one that tops my list these days is that I get timely and genuine feedback on whatever I am exploring and building. This is very important to me and it gives me a reality check. Researchers can easily get carried away with their ideas and solutions, and unfortunately, they often end up with something that does not solve something significant. I really appreciate the fact I am not doing research just for the sake of doing it but rather for building something which is going to be used.

 

 

I would advise all doctoral students to develop two habits.

1. Break your own limits and keep going into the depth of your topic as much as you can. This is definitely going to bring you excellence.

2. Take some time out, and by thinking from a very high level try to put your work into real word contexts. These reflections will help you to see the global picture, and later connect to different career paths after your thesis.