Yuvaraj Govindarajulu
About
Head of Research - AIShield, Bosch Global Software Technologies, Bengaluru (India)
Focus Topics: AI Security, AI RedTeaming, AI Trustworthiness.
Education
Master of Science (INFOTECH - Embedded Systems), University of Stuttgart, Stuttgart (Germany), with focus on Artificial Intelligence.
Previous Positions
- 2020-2021, Embedded Software Engineer (HMI & Secure Bootloader), Lorch Schweisstechnik, Auenwald (Germany)
- 2019-2020, Tutor, Deep Learning - Master Laboratory course, Institute of Signals and System Theory, University of Stuttgart (Germany)
- 2017-2019, Working Student - Embedded Prototyping for Automotive ECUs, Robert Bosch GmbH, Schwieberdingen (Germany)
- 2012-2017, Senior Software Engineer - Complex Drivers Development - Fuel Injection Software for Automotive ECUs, Robert Bosch, Bengaluru (India)
Memberships:
- Core Author, OWASP AI Exchange (AI Red Teaming), Since May 2024
- Member, Technical Committee - Securing AI, European Telecom Standards Institute (ETSI), Since Aug 2022
- Member, WG4 - Security Controls and Services Working Group, ISO/IEC/JTC1 SC27.
- Panel Member, Panel P6 - For inputs on JTC 1/SC 27 Documents Panel, LITD 17 (Information Systems Security and Privacy), Bureau of Indian Standards ((Mirroring Committee ISO)
Contact:
Master Thesis
Title: Human Activity Recognition and Study of Dynamic Filter Networks for Position-aware detection
- Aim: To build a generic framework for Human activity recognition using smartphone Inertial sensors, use of Dynamic Filter generic networks to build a single activity recognition model for multiple positions
- Building Recurrent Neural Network models for multi-variate time series data.
- Signal pre-processing steps: Design and use of filters for noise and gravitation component removal, data normalization, sensor time-synchronization and sliding window techniques.
- Study of multi-input, multi-output neural network architectures using concepts of Dynamic Filter Networks.
- Libraries used: Tensorflow, Keras, Matplotlib, Scikit, Pandas among others.
- Use of Tensorflow profiling to visualize tasks between CPU and GPU to identify performance bottlenecks.
- Deployment of trained model on an Android smartphone using Tensorflow lite.
Links: Thesis Presentation