Ai Engineer
Current▪ Prepared and designed test cases in the Saudi dialect to evaluate Speech-to-Text models.▪ Created test cases in the Saudi dialect to evaluate Large Language Models (LLMs) such as Jais, Llama 3.1, Qwen 2.5, Mistral, and Mixtral.▪ Crafting specialized prompts to enhance LLM performance, then evaluating and comparing model output (such as Jais, Llama 3.1, Qwen 2.5, Mistral, and Mixtral.).▪ Assessing the accuracy of model outputs using tools such as BGE-M3, Paraphrase, and Ragas.▪ Developed a General Evaluation framework to evaluate LLM outputs based on a given custom criteria using the LLaMA 3.1 70B model and Prompt Engineering techniques like Chain of Thought Prompting.▪ Created prompts to strengthen LLMs against attacks and guide model responses for secure outputs.▪ Utilized LangChain's self-critique feature to create prompts to control and detect potential attacks, such as prompt injection, ensuring that model outputs are safe before being presented to users.
