At upGrad, I developed a medical image synthesis solution using ESRGAN to address healthcare data imbalance. I used exploratory data analysis to identify underrepresented skin cancer classes and generated synthetic images to improve dataset balance and model generalization.
I also developed a multi-agent virtual insurance assistant with supervisor-based orchestration and a FAISS-powered RAG pipeline. As a Freelance LLM Trainer & Evaluator at Outlier.ai, I evaluated model reasoning and factual accuracy, identified hallucinations and bias patterns, and contributed to dataset refinement and response scoring.
