LLM Fine-Tuning and AI Training (Multiple Clients, Freelance)
Led fine-tuning of Llama-3 and Mistral-7B large language models for domain-specific NLP tasks across multiple client engagements. Oversaw the AI training process to achieve GPT-3.5-level performance at reduced inference cost. Employed prompt engineering and supervised fine-tuning techniques to improve model outputs for enterprise chatbot and real-time fraud detection systems. • Applied LoRA/QLoRA fine-tuning methodology on text data for custom model adaptation. • Integrated and assessed model updates using open-source evaluation frameworks and user feedback loops. • Configured data pipelines for supervised learning cycles, maintaining high-quality training datasets. • Collaborated cross-functionally with product and analytics teams to meet client-specific AI objectives.