Bring a better customer experience with the latest technologies in machine learning and computer vision.
Image Recognition (2016-): We think image recognition is one of the most important technologies for us. We can reduce users tasks in our service by that, such as filling some forms automatically.
Help Center (2016): We have improved US Mercari help center based on transaction status for each item so that users find articles what they want easier. Design updates were also very successful in bringing better user experiences.
SNS & Game Data Analysis & Big Data Analysis System Operation (2012-2016): My responsibilities include SNS & game data analysis, big data analysis system operation, developing user tracking system / log collection system / BI tool / recommendation algorithms, and so on. Since the number of SNS & game data is huge like billions records per day, we have created and used on-premise Hadoop based analysis system(Hive, Presto, Spark, HBase, etc...). We use fluentd for log collector and webhdfs to store log data in hdfs. That realizes semi-realtime anaysis for data analysts. Not only system and service developing, but also I engage in KPIs and metrics design for growth hackking.
Railway Automatic Ticket Gate (2010-2011): The aims of this project are to eliminate illegal passage and to realize supporting system for disability people in a train station. For that purpose, several cameras and an image processing engine are being installed to railway automatic ticket gate. Responsibilities include project management, designing and implementing scene understanding algorithm on CPU and DSP, and making proposal the work to clients.
Image Restoration (2008-2010): This was joint work with Prof. Kawasaki at Kagoshima University. In this project, image restoration especially super-resolution and deblurring have been studied. Responsibilities include designing and implementing image restoration algorithm. The implementation works on not only CPU, but also GPU with Nvidia CUDA. Our papers were accepted to several international conferences related to computer vision which acceptance rates were less than 30%.
Railway Fare Calculation (2006-2008): This was joint project with Prof. Ikegami at Seikei University. Responsibilities include designing and implementing efficient railway fare calculation algorithm. The software has been used as a core engine for fare calculation in Suica/Pasmo area. And that can reduce computational time for fare calculation and time and effort for maintaining the fare calculation system dramatically. My company and Scheduling Society of Japan rewarded me for my achievement.
Image and Voice Understanding (2004-2006): Main responsibility was to improve the performances of OCR(Optical Character Recognition) engine and singing voice classification engine. The contributions were reducing computation time for OCR by a coarse-to-fine approach using multiple OCR engines hierarchically, and improving the performance of singing voice classification using a reliability parameter which was measured by my proposed method.