Sushovan Jena Email & Phone Number

As a Graduate Research Assistant at Hitachi in collaboration with IIT Mandi, I am working on Anomaly Segmentation of Industrial Objects on Edge Devices using Deep Learning Model Compression techniques. My research aims to improve the Generalisation of Unsupervised methods using Knowledge Distillation and Quantization of Deep Learning models with Latency constraints. Implemented and assessed various Teacher-Student based CNN frameworks, Generative-based methods, and Memory-Module based techniques for Unsupervised Anomaly Segmentation, and leveraged Attention mechanism, Knowledge Distillation, Post-Training Quantization, and Quantization-aware Training across PyTorch and TensorRT frameworks to reduce latency on resource-constrained devices. I have also deployed models on NVIDIA Jetson Xavier NX.Papers accepted - Distributed Convolutional Attention (DCAM) based strategy to improve Knowledge-Distillation eliminating Cross-class interference at International Conference on Pattern Recognition (ICPR) 2024.Multimodal English-Low Resource Language (Hindi, Malayalam, Bengali, Hausa) Translation on Mistral 7B and PaliGemma 3B at WMT workshop in EMNLP 2024.Prior to this, I worked on Data Engineering and Machine Learning on SAP, and obtained the SAP Certified Development Specialist - ABAP for SAP HANA 2.0 certification. I also completed the Convolutional Neural Networks from One Fourth Labs, IIT Madras.I am passionate about applying Generative AI, Machine Learning and Deep Learning to solve real-world problems, especially in the domains of Computer Vision, NLP and Edge Computing.I have an analytic view, an intuitive mind, and an Indian wisdom that guide me in my research and learning endeavors.