Prathik Kodanda Murthy

Research developer for state of the art implementation of attention-based transformers for motion trajectory prediction in autonomous vehicles

Devised a state-of-the-art deep learning framework to generate and augment the authenticity of simulated ultrasonic echo data, applying a novel style transfer approach rooted in Cycle consistent Generative Adversarial Networks principles, targeted for validation of automated parking functionalities. || Successfully implemented and deployed a novel Generative Adversarial Network (GAN) architecture tailored for a zero-inflated dataset, characterized by a complicated ~2% non-zero irregular timeseries peaks.

• Collaborated and Implemented Convolutional Neural Networks (CNNs) with PySpark in Azure Databricks to measure AdBlue solution quality in exhaust systems. Investigated with Azure AutoML. || • Authored a novel data cleaning for improved quality and training pipeline script in Python, widely used within Bosch.

Engineered 98% accurate Tampering Detection algorithm for powertrain sensors using LSTMs, GRUs, and Autoencoders, deployed with Azure and MLOps. |Collaborated to secure models against attacks. The AI models were embedded and deployed for real-time systems.| Led T.A.R.S application development using Angular and Docker. || Developed RTOS-based embedded software, testing device drivers and sensors for automotive OEMs.| Automated IoT lab control for remote ECU testing using Raspberry-Pi, creating a GUI in RBIN.