BRANNON KRISTION

AI Training Specialist and Technical Evaluator with extensive experience in evaluating AI-generated responses for accuracy, relevance, and compliance. Demonstrates proficiency in software development and technical quality assurance, including code review, debugging, and structured output validation. Skilled in data annotation, content review, and quality control to ensure high-quality deliverables. Expertise in AI alignment knowledge such as RLHF and supervised fine-tuning. Proficient in programming languages including Python, C++, JavaScript, and development tools like Git and Docker. Career goals include advancing AI safety principles while enhancing technical evaluation processes.

• Large Language Model Evaluation
• Assess responses for accuracy, relevance, clarity, completeness, and alignment with prompts.
• Compare alternative answers using consistent evaluation criteria.
• Identify misunderstandings, missing information, unsupported claims, and irrelevant content.
• Distinguish between fluent responses and responses that are accurate and useful.
• Provide clear rationales supporting evaluation decisions.
• Reinforcement Learning from Human Feedback (RLHF)
• Understand how human feedback can help improve AI-generated responses.
• Recognize how response comparisons and preference ratings can inform model training.
• Apply objective criteria when comparing model outputs.
• Understand the importance of consistent judgments and clear evaluation rationales.
• Supervised Fine-Tuning and Data Annotation
• Understand how high-quality examples support supervised model training.
• Recognize the importance of accurate labels, consistent formatting, and clear instructions.
• Review annotated examples for completeness and alignment with guidelines.
• Identify inconsistent labels and ambiguous instructions that may reduce data quality.
• Technical and Code Evaluation
• Review technical answers for logical correctness, readability, relevance, and completeness.
• Identify potential syntax errors and flawed reasoning in familiar programming languages.
• Assess whether technical explanations accurately describe the presented code.
• Recognize the importance of testing, reproducibility, and evidence.
• AI Safety and Responsible Evaluation
• Identify potentially misleading, biased, unsafe, or unsupported model outputs.
• Assess responses against applicable task-specific safety requirements.
• Recognize the importance of privacy and responsible data handling.
• Document concerns clearly and follow applicable escalation procedures.

• Areas of Interest: AI training, LLM evaluation, AI response quality analysis, AI code evaluation, data annotation, RLHF, SFT data review, and AI safety evaluation.
• Languages: English (Native), Proficient in German and Spanish