Lalitha Guru Swaminathan

Built cloud-based NLP pipelines using Azure AI, Github, and Copilot Studio for automated document summarization and semantic search. Developed and deployed AI workflows on AWS integrating large-scale data pipelines and ML models for enterprise clients. Applied RAG techniques with JSON NoSQL-style datasets in Azure AI and integrated LLM APIs into business workflows. Collaborated within Agile teams globally to ensure secure and scalable solutions meeting tight deadlines.

Designed and deployed an English–Malay LLM-based translation model improving accuracy by 18% via contextual tuning. Optimized PyTorch computer vision models for energy monitoring by addressing latency and reproducibility issues in cloud environments. Led AI workflow demonstrations, gathered client requirements, and supported workflow adoption. Gained familiarity with MLOps (MLflow) and collaborated with Agile teams across three countries, cutting validation time by 30%.

Designed SEEA-UNet, a novel hybrid model enhancing brain tumor segmentation using refined spatial and channel attention mechanisms. Verified and preprocessed over 1,200 annotated MRI images resolving inconsistencies and reducing segmentation error by 21%. Optimized hybrid U-Net in PyTorch achieving 0.0112 validation loss within 10 epochs to accelerate convergence and improve precision. Published research on arXiv and presented findings simplifying complex model mechanics for diverse audiences.