aadil rasheed

I’m an AI/ML Engineer and Data Scientist focused on end-to-end machine learning—turning data into production-ready NLP, LLM, and speech AI systems. I enjoy designing multilingual solutions, optimizing models for real-world performance, and building reliable deployment pipelines.

At Nucleosight, I fine-tune and deploy transformer-based LLMs for domain-specific conversational AI, including multilingual systems for Urdu, Kashmiri, and English. I’ve built ASR solutions using Whisper, ESPnet, Kaldi, and Hugging Face Transformers, and developed neural TTS pipelines for low-resource and underrepresented languages. I also lead large-scale dataset collection and QA, and implement RAG-based conversational AI systems with grounded, factual responses.

Earlier, as a Machine Learning Researcher at the University of Kashmir, I developed deep learning frameworks for multimodal representation learning using hierarchical autoencoders. I designed and trained graph neural networks for complex network analysis, and created experimental pipelines for training, hyperparameter tuning, validation, and evaluation. My research work spans uncertainty-aware deep learning for clustering and peer-reviewed publications on deep learning architectures and multimodal learning.

I bring a strong toolkit across deep learning, generative AI, model optimization, and data engineering—from dataset creation and annotation to evaluation and cloud deployment on Azure, AWS, and GPU environments. I’m motivated by building systems that are not just accurate, but dependable, scalable, and usable by teams in production.