Winner. Scientist of the Year 2022

Sumedh Datar

Academic title, degree: M.S. in Computer Science
Fields of science: Computer Science
Research interest: Computer Vision, Deep Learning
Career Development: Experienced
Organization: 7-Eleven
Position: Senior Machine Learning Engineer
City: Grapevine, Texas
Country: USA

Sumedh Datar has a strong background in Full-Stack Development and gained extensive experience developing end-to-end systems. He has demonstrated his expertise by creating systems encompassing the front-end, back-end, middleware, database, and machine learning components. This unique combination of skills positions Mr. Datar as an engineer who works at the intersection of engineering, research, and product development.

He obtained Bachelor of Engineering in Medical Electronics, specializing in Medical Imaging and Programming. During his undergraduate studies, Sumedh worked on developing an algorithm to identify cancerous regions from endoscopic images. The project he worked on received third place across the department.

After postgraduate studies, Sumedh worked for two startups before coming to the US to obtain his Master's. At the first startup, he performed research in identifying non-invasive ways to detect painless oral cancerous regions. This helped patients to go for diagnosis without any pain or discomfort freely. The solution he developed was very cheap and affordable, it could be used anywhere, and thus it was found to be very useful in areas where healthcare accessibility is rare.

Mr. Datar has also conducted research in the field of face recognition using Deep Learning and has successfully developed a complete product around it. The automated attendance management system he created tracks attendance, generates reports, and is used by thousands of people. This system has effectively solved the proxy attendance problem and saved significant amounts of money for various skill development organizations.

Later, he performed original research applying Deep Learning to identify cancerous regions from pathology images. The product is used by more than 500 patients and provides results in a few hours compared to a few weeks with manual techniques.

Sumedh Datar also performed research in asset protection and was the first researcher to apply Deep Learning-based object localization to a real-world problem. The project's success led to expanding the business and tapping into new areas.

Mr. Datar has made significant contributions to the field of cashier-less stores and has been awarded several patents for his work. The cashier-less store involves complex components such as sensor fusion, distributed computing, computer vision, hardware, and machine learning, making it challenging to conduct research and develop a solution. Despite these difficulties, Mr. Datar has demonstrated his expertise by efficiently creating optimized solutions to run a cashier-less store with reduced maintenance requirements.

Currently, he works on applied Deep Learning in retail, and has 4 granted patents and more than 10 applied patents in the fields of Computer Vision, System Design, and Deep Learning. He is part of the leadership team where he mentors his peers and also gives direction to project. He is also deeply involved in architecting scalable systems and defining standardized practices for the department.

During mentoring, Mr. Datar collaborated on continuous research on identifying cancerous regions from endoscopic images. The students won third prize, and together they discovered that there were not many open-source datasets related to endoscopic images. Sumedh has created a platform and open-sourced the dataset for people worldwide to use and contribute to the betterment of society.

Mr. Datar is an exceptional researcher who not only conducts research but also transforms these research prototypes into practical solutions that have made a significant impact on the lives of thousands of people worldwide. His work has been recognized, and Sumedh received multiple corporate awards.

Apart from work, he is an invite-only member of the Linux Foundation, and top academic institutes around the world are using the open-source contribution that he has made. He designed coursework in Computer Vision and Deep Learning that had a significant impact, with his modules being used by 10 of the top academic institutions worldwide.

His expertise in the field of Computer Vision is further demonstrated by his role as a peer reviewer for the British Machine Vision Conference, one of the top computer vision conferences in the world. Sumedh has also been called upon to review several technical books related to Machine Learning Explainability and Scaling Machine Learning Systems.

He has had several speaking engagements as an expert in the field, demonstrating his knowledge and credibility to the AI community. He was invited to be interviewed for AI Time Journal and as a panelist to speak alongside other industry leaders about operationalizing Deep Learning Computer Vision-based models for production. In addition, Mr. Datar gave an interview for a podcast where he discussed the role of AI in business and the importance of incorporating AI into organizational strategy, and a talk on YouTube on how to pursue a career in AI, Computer Vision, and Deep Learning.

His work was recognized as outstanding by the Institute of Data Scientists and Analysts, and Scholarly Academic and Scientific Society, and he received fellowships from these institutes. As part of governance, Sumedh Datar is on the Editorial Board for the popular journal IARJET. In recognition of his leadership skills and contribution to society, Mr. Datar was also invited to join the Harvard Leadership Excellence Community.

With his thorough knowledge in the Computer Vision field, Sumedh was invited to publish his article on Analytics Vidhya (128.9k daily visitors). He also contributed with the article "Application of Pose Estimation with Deep Learning" to DZone (152.4k daily visitors).

His contribution to improving healthcare by using AI, original invention for various deep learning-based techniques with patents in retail, and his open source contribution shows Sumedh Datar combines both a passion for research with a desire to make a real-world impact and is one of the few has risen to the very top in his field.