Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Hello! I'm Shreyasi Ghosh, an AI/ML researcher and software developer currently pursuing a Master's in AI. I work on large-scale data processing, transformer-based models, and simulation-driven ML, aiming to translate research into practical tools for industry. I'm passionate about end-to-end ML workflows—from data engineering and model training to visualization and deployment. I enjoy collaborating across disciplines and continually exploring novel approaches to improve accuracy, efficiency, and real-world impact.…Hello! I'm Shreyasi Ghosh, an AI/ML researcher and software developer currently pursuing a Master's in AI. I work on large-scale data processing, transformer-based models, and simulation-driven ML, aiming to translate research into practical tools for industry. I'm passionate about end-to-end ML workflows—from data engineering and model training to visualization and deployment. I enjoy collaborating across disciplines and continually exploring novel approaches to improve accuracy, efficiency, and real-world impact.

SHREYASI GHOSH

Data Scientist, AI Developer, Full Stack Developer, +2





Hello! I'm Shreyasi Ghosh, an AI/ML researcher and software developer currently pursuing a Master's in AI. I work on large-scale data processing, transformer-based models, and simulation-driven ML, aiming to translate research into practical tools for industry. I'm passionate about end-to-end ML workflows—from data engineering and model training to visualization and deployment. I enjoy collaborating across disciplines and continually exploring novel approaches to improve accuracy, efficiency, and real-world impact.…Hello! I'm Shreyasi Ghosh, an AI/ML researcher and software developer currently pursuing a Master's in AI. I work on large-scale data processing, transformer-based models, and simulation-driven ML, aiming to translate research into practical tools for industry. I'm passionate about end-to-end ML workflows—from data engineering and model training to visualization and deployment. I enjoy collaborating across disciplines and continually exploring novel approaches to improve accuracy, efficiency, and real-world impact.

Available to hire

Hello! I’m Shreyasi Ghosh, an AI/ML researcher and software developer currently pursuing a Master’s in AI. I work on large-scale data processing, transformer-based models, and simulation-driven ML, aiming to translate research into practical tools for industry.

I’m passionate about end-to-end ML workflows—from data engineering and model training to visualization and deployment. I enjoy collaborating across disciplines and continually exploring novel approaches to improve accuracy, efficiency, and real-world impact.

Skills

Experience Level

Expert

Expert

Expert

Expert

Intermediate

Language

English

Fluent

German

Advanced

Work Experience

AI Intern at AUDI AG

August 1, 2025 - Present

Research with Databricks and PySpark for large-scale vehicle log data processing, including data engineering and model prototyping in Python. Collaborated across teams to enable scalable analytics workflows for automotive data.

Software Developer (Student Research Assistant) at Fraunhofer ITWM

July 31, 2025 - October 6, 2025

Developed a Python script to generate a repeatable grid-like textile structure applying Dirichlet boundary conditions to study strain distribution. Implemented a Transformer-based anomaly detection model to identify inconsistencies in simulation outputs, improving error detection in textile simulations by 30%. Used ViT for automated fiber defect classification, reducing manual inspection by 35%. Created Python-based cloth simulation software with physics-informed ML (PINNs) to enhance fabric behavior modeling, achieving ~30% accuracy improvement.

Master Thesis Student at Rheinland-Pfälzische Technische Universität Kaiserslautern - Robotics Research Lab (RR Lab)

May 31, 2025 - October 6, 2025

Developed a novel framework, LRDG, to explicitly remove domain-specific features for improved model generalization across unseen domains. Trained deep learning models with ResNet backbone, achieving 15-20% improvement on PACS. Integrated Detectron2 for domain-specific feature removal, boosting cross-domain performance by ~20%. Conducted experiments on multispectral and amplitude-phase recombination image datasets using PyTorch, OpenCV, and GANs.

Software Developer (Student Research Assistant) at FBK - Rheinland-Pfälzische Technische Universität Kaiserslautern

July 31, 2025 - October 6, 2025

Developed Python-based software demonstrator for predicting manufacturing processes, increasing analysis efficiency by 40% through automation. Designed and implemented a 3D visualization tool using PythonOCC and PyQt5, enhancing real-time model interaction and usability by 35%. Explored Graph Neural Networks to improve feature extraction and classification, improving manufacturing data analysis accuracy by 20%.

Software Developer (Student Research Assistant) at Refactum

December 31, 2024 - October 6, 2025

Developed a Web API with FastAPI, incorporating ML-powered recommendation systems for intelligent CAD model retrieval based on past user interactions, reducing response times by 30%. Rendered 3D CAD models using PythonOCC and integrated deep learning-based mesh optimization techniques to improve model accuracy and reduce rendering time by 25%. Integrated Neo4j for efficient data storage and retrieval, optimizing query speeds by 40% and improving scalability. Implemented robust unit and integration tests using Pytest, increasing code reliability.