At Hindustan Aeronautics Limited (HAL), I developed backend components for a RAG-based Fake News Detection system, integrating retrieval from Wikipedia and the GNews API with an open-source LLM for claim extraction and classification. I also implemented its end-to-end pipeline as a modular REST API.
I engineered the Expense Tracker backend with Java and Spring Boot, and built a YouTube sentiment analysis API that classified comments and reduced response latency from 20–60 seconds to under 100 milliseconds. In a published comparative study, I evaluated ResNet18 and ResNet50 for brain MRI classification and recommended the lighter model for resource-constrained clinical settings.
