Mahalakshmi Manoharan

I’m a Research Analyst and LLM Specialist with advanced expertise in mathematics, AI research, and large language model training and evaluation. I focus on building and assessing LLMs for complex problem-solving across algebra, calculus, geometry, and Olympiad-level mathematics.

In recent roles, I’ve built and evaluated LLMs and GenAI outputs for STEM reasoning, performing dataset annotation and curation across image, video, and text. I extract and structure information from PDFs and documents to support AI data pipelines, then apply quality validation to strengthen accuracy, reliability, and reasoning performance.

I also train and evaluate LLMs for advanced mathematical reasoning tasks—prompt engineering, data pre-processing, and quality audits included. My evaluation approach uses structured metrics such as accuracy, hallucination rate, relevance, and logical consistency to ensure outputs are not only correct, but internally coherent.

Previously, I developed and fine-tuned LLMs using Python and PyTorch for multi-step and competition-level math reasoning, applying techniques like SFT, RLHF, and RLMF. I bring an analytical, error-driven mindset—conducting error analysis and root-cause debugging—along with a strong background in mathematics instruction and learning-focused content development.