LLM-Powered Synthetic Exam Question Generation Engine (Independent AI Systems Engineer)
Built a synthetic MCQ generation engine leveraging few-shot learning and LLMs for exam preparation. Extracted question schema from structured PDFs and trained the system to mimic authentic exam patterns. Incorporated user feedback to iteratively improve model outputs and reduce hallucination. • Developed PDF parsing and structure extraction logic. • Tuned few-shot LLM prompts based on user-corrected feedback. • Enhanced data quality by tracking and using user input corrections. • Utilized OpenAI API and deployed continuous feedback optimization.