Fine-Tuning Engineer (LLM QA Bot)
Fine-tuned the Mistral 7B large language model using QLoRA for a conversational question answering bot project. The experience involved adapting model weights via supervised learning on QA datasets and deploying the resultant model to production. Tasks included dataset curation, hyperparameter tuning, and evaluating bot response accuracy. • Achieved approximately 90% response accuracy on validation datasets. • Automated and monitored inference on AWS SageMaker. • Reduced inference latency by 20% through optimization steps. • Collaborated closely with cross-functional teams on model assessment.