Darshan Premraj

I’m an AI Model Trainer & GenAI Engineer with hands-on experience training, evaluating, and red-teaming large language models, backed by data science and machine learning engineering fundamentals. I focus on high-integrity human feedback by combining LLM evaluation (Ragas, TruLens) with statistical methodology like hypothesis testing, A/B testing, and inter-annotator agreement.

In my recent role as a Data Automation Engineer, I engineered automated data pipelines that reduced manual processing time by 30% and improved accuracy across 4+ institutional departments. I also integrated LLM API calls (OpenAI) into reporting workflows for natural-language operational analytics, and prototyped an LLM-assisted data categorization tool using LangChain with validation against ground-truth labels.

Across projects, I build systems that are both technically rigorous and deployable—RAG pipeline design, prompt engineering, structured-output validation, and containerized services (Docker, FastAPI). I’ve applied advanced neural architectures (Physics-Informed Neural Networks, Kolmogorov-Arnold Networks) and classical ML to benchmark performance, track experiments in MLflow/Weights & Biases, and deliver reproducible, quality-focused results.