Nirmal Antony

Developed a generative image-to-image translation pipeline to improve autonomous driving perception under adverse weather conditions. I designed and benchmarked diffusion models, GANs, CNNs and vision-language architectures using the ACDC dataset and multiple publicly available weather datasets. My work included building scalable data pipelines to handle diverse weather scenarios, fine-tuning models for specific conditions, and integrating semantic priors to enhance consistency.

Conduct research on multiword expressions (MWEs) and their impact on syntactic and semantic parsing within the DFG-funded MWE-SemPrE project by developing and testing BERT models, applying machine learning and AI techniques to enhance NLP system performance on linguistic datasets, and contributing to the integration of linguistic theory with computational methods in computational linguistics.