Hemanth Volikatla is a Senior Technical Service Manager at SAP with over 20 years of experience in IT, specializing in AI, cloud computing, and digital transformation. He has played a pivotal role in optimizing business operations and enhancing global health initiatives, particularly in the healthcare sector, through his expertise in cloud technologies like SAP HANA, AWS, Azure, and Google Cloud. Hemanth has extensive customer advisory experience, having worked closely with SAP ECS customers as well as global clients across industries, offering expert guidance on cloud strategies, AI implementations, and system optimizations. Hemanth has enormous experience in research, particularly in AI, machine learning, and cloud technologies. His work has led to numerous publications, and he has contributed significantly to optimizing sensor data and developing algorithms to improve real-time and post-activity correction in fitness tracking systems. His research has shaped innovative solutions in sensor fusion, predictive analytics, and data processing, enhancing technologies in a range of industries, including healthcare and fitness. Hemanth has extensive experience in customer-facing roles, particularly with SAP ECS (Enterprise Cloud Services), where he has played a key role in guiding and mentoring customers throughout their digital transformation journeys. In his capacity, he has been instrumental in safeguarding customer investments by ensuring the successful implementation of cloud solutions and offering continuous support to help them navigate and optimize their cloud interfaces. He has provided expert guidance on cloud strategies, particularly in the areas of data analytics, sensor fusion, and AI-driven solutions, ensuring customers achieve their business objectives efficiently and effectively. Before his tenure at SAP, Hemanth made significant contributions during his time at Hewlett Packard (HP), where he led a joint HP-SAP project, resulting in a published paper on the deployment of Service Registries. This achievement underscored his ability to bridge gaps between major organizations and deliver impactful results. Throughout his career, Hemanth has been deeply committed to social causes, dedicating time to mentoring engineers and professionals, particularly those pursuing careers in AI and machine learning. His efforts in driving digital equity and empowering future generations of technologists have been a cornerstone of his career. Hemanth is also a respected thought leader and has contributed to various peer reviews, judging roles, and research projects. His expertise is regularly sought after in industry forums, where he assesses innovative technological solutions. He is continuously exploring the intersection of AI, cloud technologies, and real-time data processing to drive forward-thinking solutions in both the tech and business sectors. His contributions have earned him numerous prestigious awards, including the Stevie International Business Award for Technology Excellence – Lifetime Achievement Award: Celebrating his unparalleled contributions to enterprise IT innovation., Globee Awards: Recognized for advancements in AI/ML – Executive of the Year | Artificial Intelligence and IT Professional of the Year – IT transformation and supply chain optimization., and the IEEE Outstanding Engineer Award: Honoring his extraordinary engineering achievements in the IT domain, underscoring his impact in the AI and technology fields. Additionally, Hemanth is the author of Practical Guide to Cloud-Based Machine Learning: Deploying AI Models in the Cloud, a highly regarded resource for professionals looking to deploy AI in cloud environments, and a valuable source for advancing cloud-based machine learning applications.
AI/ML, SAP, Cloud Computing
This study addresses the challenge of improving accuracy in fitness tracking data, particularly when using GPS and high-frequency IMU sensors. GPS data is often noisy, leading to issues such as false paths, where users appear to have traveled through buildings or bodies of water. To tackle this, the study proposes using accelerometers, gyroscopes, and magnetometers—components of an Inertial Measurement Unit (IMU)—to counteract inaccuracies in both high-frequency IMU sensors and low-frequency GPS data. The research explores two correction strategies: real-time correction during the activity, which demands computational efficiency for immediate data processing, and post-activity correction, which leverages both previous and future sensor data for more accurate estimations. The study compares two approaches: the Kalman Filter, a traditional method from the 1960s that performs a reverse pass post-activity to improve estimations, and Recurrent Neural Networks (RNNs), including Gated Recurrent Units (GRUs) and Long Short-Term Memory (LSTM) Networks, which use built-in memory modules to learn patterns in sensor data without requiring sensor variance information. The findings suggest that RNNs outperformed the Kalman Filter in producing more credible results, especially in handling the noisy and low-frequency nature of the data. However, inconsistent ground truth data made it difficult to perform a reliable quantitative evaluation of the models. While RNNs showed promise due to their ability to remember long-term dependencies and learn from sequential data, the Kalman Filter still provided value, particularly with its post-activity reverse pass. The study concludes that further research is needed to address the issue of ground truth data inconsistency and to explore hybrid approaches that combine the strengths of both methods. Future investigations could also focus on more robust real-world testing to improve the accuracy and reliability of these techniques for fitness-tracking applications.