Designed and built an end-to-end LLM-driven AutoML platform that converts natural language requests into executable, optimised machine learning pipelines. Integrated LLMs with AutoML frameworks to automatically generate, validate, and optimise Python ML code, enabling rapid model development with minimal manual intervention. Engineered a unified ML workflow supporting both TensorFlow/Keras and PyTorch, allowing dynamic model selection, training, evaluation, and deployment within a single pipeline. Implemented robust code validation and execution safeguards to ensure syntactic correctness, runtime stability, and reproducibility of generated ML pipelines. Developed and maintained custom datasets to improve model reliability, safety constraints, and instruction-following behaviour in automated code generation systems. Built data preprocessing, feature engineering, and dataset versioning workflows to support repeatable experimentation and model improvement. Implemented model evaluation, hyperparameter optimisation, and automated benchmarking across multiple architectures using AutoML techniques. Performed large-scale experimentation to compare model performance, interpretability, and stability under real-world usage scenarios. Applied MLOps best practices, including modular pipeline design, experiment tracking, automated testing, and model performance monitoring. Focused on building scalable, production-ready ML systems with clear failure handling, observability, and maintainability.