Nikhil Kumar Singh

I’m a Jr Project Engineer at AmritaCREATE Lab, focused on designing and developing LLM- and VLM-based solutions for assistive communication, sign language translation, and multilingual educational tools. I integrate speech, text, visual, and multimodal data to build inclusive AI platforms that support accessibility in education and healthcare.

Earlier, as a Research Assistant Intern, I engineered Variational Autoencoder (VAE) models for synthetic patient data generation, trained on 10,000+ EHR samples with 85% reconstruction accuracy. I also integrated privacy-preserving mechanisms to maintain over 95% synthetic data utility while preventing patient re-identification, and contributed to a research paper currently under clinical review.

My work spans practical MLOps and model deployment thinking, from transformer-based news summarization to secure, scalable synthetic patient modeling. Through projects like AnonyGraph (hybrid VAE-GNN using medical knowledge graphs) and full-stack systems like InsightInk, I consistently aim to turn research into reliable, usable tools.