Summary
Highly accomplished AI Trainer with 6 years of experience specializing in developing and refining AI models through meticulous data annotation, robust quality assurance, and iterative feedback cycles. Proven ability to significantly improve model accuracy, reduce hallucination rates, and enhance user experience across diverse natural language processing and computer vision applications. Adept at leading cross-functional teams and implementing scalable training methodologies to achieve ambitious project goals.
Experience
AI Training LeadAnthropic
- Led a team of 10 AI Trainers to improve the safety and helpfulness of large language models, resulting in a 15% reduction in harmful outputs.
- Designed and implemented new annotation guidelines, increasing data consistency by 20% and accelerating model iteration cycles by 10%.
- Managed feedback loops with ML engineers, providing critical insights that led to a 5% improvement in model benchmark scores.
- Mentored junior trainers, streamlining onboarding processes and reducing ramp-up time for new hires by 25%.
Senior AI TrainerScale AI
- Annotated and validated over 100,000 data points for NLP and computer vision projects, contributing to a 90%+ data quality rating.
- Developed internal tooling and scripts that automated repetitive tasks, saving the team an estimated 15 hours per week.
- Collaborated with ML engineers to identify and resolve model biases, improving fairness metrics by an average of 12% across multiple datasets.
- Conducted quality assurance checks on external vendor data, reducing error rates by 30% and improving overall project timelines.
Projects
Prompt Optimization for E-commerce Chatbot
- Developed and iteratively refined 200+ prompts for an e-commerce chatbot, improving customer query resolution rate by 18%.
- Conducted A/B testing on prompt variations, identifying optimal strategies that reduced chatbot hallucination errors by 10%.
- Documented a comprehensive prompt engineering guide, enabling future development teams to maintain high performance.
Custom Data Annotation Platform for Medical Imaging
- Designed and implemented a web-based tool for medical image annotation, reducing labeling time by 25% for complex tasks.
- Integrated AI-assisted pre-labeling features, which accelerated the annotation process by an additional 15%.
- Enabled multi-user collaboration and real-time QA, increasing annotation consistency by 20% across diverse datasets.
Bias Detection and Mitigation in Language Datasets
- Developed Python scripts to identify and quantify gender and racial biases in publicly available NLP training datasets.
- Proposed and implemented data augmentation and re-weighting strategies that reduced detected biases by up to 30%.
- Published findings on a personal blog, contributing to broader discussions on ethical AI development and fair data practices.
Education
University of California, BerkeleyMaster of Science in Data Science
- Achieved a GPA of 3.8/4.0, specializing in Machine Learning and Natural Language Processing.
- Coursework included Deep Learning, Statistical Modeling, and AI Ethics.
- Developed an NLP sentiment analysis model for customer reviews as part of capstone project.
University of California, San DiegoBachelor of Science in Computer Science
- Graduated Cum Laude with a GPA of 3.7/4.0.
- Honors Thesis: 'Optimizing Data Labeling Workflows for Enhanced Model Performance'.
Skills
AI/ML Tools
TensorFlowPyTorchHugging FaceLangChainOpenAI APIWeights & Biases
Data Annotation & Quality
Label StudioProdigyData ValidationGuideline DevelopmentQuality AssuranceInter-Annotator Agreement
Programming & Scripting
PythonSQLBashJupyter NotebooksPandasNumPy
NLP & NLU
Large Language Models (LLMs)Prompt EngineeringText ClassificationNamed Entity RecognitionSentiment AnalysisGenerative AI
Project Management & Collaboration
JiraConfluenceAgile MethodologiesTeam LeadershipCross-functional CollaborationStakeholder Communication