Job Summary:
TSMC Arizona offers an opportunity to work at the most advanced semiconductor fab in the United States. As a Senior AI Model Fine-Tuning Engineer, you will adjust the behavior and functional abilities of AI models to enhance their adaptability and intelligence, utilizing advanced techniques like prompt engineering and RLHF.
Responsibilities:
• Lead the fine-tuning process for large pre-trained models, focusing on making models behave appropriately in different contexts (e.g., following instructions, answering questions, or performing tasks).
• Design and implement prompt engineering strategies to help the model produce more accurate, relevant, and coherent outputs.
• Apply Reinforcement Learning from Human Feedback (RLHF) and other behavioral fine-tuning methods to improve the model's alignment with user needs and ethical standards.
• Collaborate with data teams to integrate relevant data and continuously improve model behavior.
• Conduct model evaluations using various performance metrics (accuracy, bias detection, user feedback) to identify areas for improvement.
• Iterate and experiment with different fine-tuning methods to achieve optimal performance for specific use cases.
• Monitor model drift and ensure that models remain consistent, reliable, and safe over time.
Qualifications:
Required:
• Bachelor's degree in Computer Science, Data Science, or a related field.
• 5+ years of experience working on fine-tuning large-scale models such as GPT, T5, or BERT, with a strong focus on behavior and functionality.
• Expertise in advanced tuning methods such as RLHF, prompt engineering, and zero-shot learning.
• Experience with popular transformer architectures and frameworks like Hugging Face, TensorFlow, or PyTorch.
• Deep understanding of LLM behaviors, including instruction-following, task completion, and ethical considerations in output.
• Proficiency in Python and experience with libraries for model fine-tuning (e.g., Transformers, DeepSpeed).
• Experience in evaluating model performance, including using metrics like BLEU, ROUGE, perplexity, and custom evaluation frameworks.
• Candidates must be willing and able to work on-site at our Phoenix Arizona facility.
• Communication
• Computer proficiency
• Presentation skills
• Listening
• Teamwork
Preferred:
• Experience with ethical AI and safety considerations, such as minimizing bias and handling adversarial inputs.
• Experience with model deployment and real-time experimentation (A/B testing).
Company:
Established in 1987, TSMC is the world's first dedicated semiconductor foundry. Founded in 1987, the company is headquartered in Hsinchu, TWN, with a team of 10001+ employees. The company is currently Late Stage.