Activity Recognition’s Medical Artificial Intelligent Assistant (MAIA) goal is to improve independent living while reducing the Japanese public healthcare cost. We develop state-of-the-art artificial intelligence methods capable of generating optimal treatments tailored to the patient’s Electronic Health Records.The project has been featured at several media press and has been receiving continuous support of the Japanese government.Latest media releases: https://www.forbes.com/sites/japan/2018/11/12/why-japans-aging-population-is-an-investment-opportunity/#3f https://tech.nikkeibp.co.jp/dm/atcl/event/15/101000171/102100008/?ST=healthhttps://japan.cnet.com/release/30245847/MAIA platform: http://maia.care/Role responsibilities and achievements:· Recruit and lead a multidisciplinary team of 8 members including Full-Stack Engineers, Machine Learning Researchers and Engineers, Data Analysts and Functional Analysts· Lead the full-stack development of the EHR management platform, the treatment recommendation and the health forecasting platform· Develop the ML auto-scalable production infrastructure in AWS used to provide access to thousands of users to more than 30 different AI models· Develop the data pipeline that includes data extraction, data visualization, data curation,ML and statistical experimentation, training/testing/validation, model debugging and production deployment of recommendation and forecasting models· Define and validate a healthcare personal data handling protocol that satisfies all safety and integrity requirements imposed by the Japanese Government to foreign companies· Assembly and configure machine learning experimentation infrastructure using high powered NVIDIA GPU machinesSide projects:· Seniors’ remote monitor device for early detection of risks· Dataset: https://github.com/activityrecognition/ARTraining· App: Thermix: Detection of elderly falls and other daily living activities using Flir thermal sensor

References:Guido Pusiol (Stanford University) http://vision.stanford.edu/pac/index.htmlZelun Luo (Stanford University) http://alan.vision/Jun-Ting Hsieh (Stanford University)Construction of Thermset, the first open for research dataset containing long term thermalvideos of seniors living independently at home or nursing homes. More info at http://activityrecognition.com/thermsetProvide technical support to the AI and medical researchers, members of the Partnership in AI-Assisted Care (PAC) team (https://med.stanford.edu/cerc/research/new-pac.html#senior), in the design, development and deployment of artificial intelligence solutions for elderly care.Participate in two publications done by the PAC team as part of the Intelligent Senior Well-Being project:· Computer Vision-Based Descriptive Analytics of Seniors Daily Activities for Long-Term Health Moni- toring.Zelun Luo, Jun-Ting Hsieh, Niranjan Balachandar, Serena Yeung, Guido Pusiol, Jay Luxenberg, Grace Li, Li-Jia Li, N. Lance Downing, Arnold Milstein, Li Fei-Fei (Published at the Machine Learning for Healthcare Conference 2018)· Computer Vision-based Approach to Maintain Independent Living for Seniors.Zelun Luo, Alisha Rege, Guido Pusiol, Arnold Milstein, Li Fei-Fei, N. Lance Downing (Published at the American Medical Informatics Association 2017)

- Led the development team of the new user application of Spinlock's NMR spectrometers and flow meters, known as NMR KERNEL (C#, WPF).- Led the development team which was responsible for the maintenance of the commercial desktop applications for the determination of the contents of Oil, Moisture and Fatty Acids in food (C#, .Net Framework 4.0).- Led the development team which designed and developed desktop applications for the flow loop operation (C#, .Net Framework 4.0).

- Designed, implemented and tested the Spinlock’s commercial desktop application, called SLK NEO MULTIEXP, for fitting low field NMR decays with exponential curves using two different methods (C#, .Net Framework 4.0).- Designed, implemented and tested the Spinlock’s commercial desktop application, called SLK CONDOR LITE, for programming and executing five different NMR experiments: Larmor Frequency Adjust, P90 and P180 estimation, T2 estimation, T1 estimation and T1-T2 estimation (C#, .Net Framework 4.0).- Designed, implemented and tested the Spinlock’s R&D platform, called SLK MAGLAB, for designing, coding and executing a vast array of NMR experiments based on the NMR_CORE data model allowing the execution of internal processes over SciLab platform (C#, SciLab, .Net Framework 4.0).

- Implemented and tested an algorithm for exponential curve fit of relaxation data using the Nelder-Mead simplex algorithm to optimize the non-linear parameters (C#).- Implemented and tested an algorithm for the reconstruction of NMR T1-T2 spectra based on the maximum entropy method (C#).- Implemented and tested a five steps algorithm for preprocessing signals of low field NMR spectrometers (C#).- Optimized and extended the functionality of a sequence pulses compiler for Spinlock’s low field NMR spectrometers (C#).- Designed, implemented and tested a communication protocol for the Spinlock’s low field NMR spectrometers using the USB Microchip interface (C#, C++).- Designed a data model for low field NMR experiments representation, known as NMR_CORE,using UML and OCL standards in accordance with ANSI/SPARC data model architecture (Final work of MSCS).- Implemented and tested an algorithm for post-processing signals of NMR flow meters to determinate the flows and cuts of: water, oil and gas (C#).