AI Training & Data Labeling Experience – PureCode AI
Text
Computer Programming/CodingPureCode AI is an AI-powered code generation and developer productivity platform that enables developers to generate, modify, debug, and optimize code directly from VS Code using natural language prompts. Responsibilities included: • Prepared, reviewed, and validated text-based datasets consisting of developer prompts, code snippets, bug-fix scenarios, and expected outputs for AI model evaluation. • Performed annotation and quality assessment of AI-generated code responses by evaluating correctness, functionality, security, performance, and adherence to coding standards. • Created benchmark datasets covering JavaScript, TypeScript, React, Next.js, Node.js, Python, Java, cloud infrastructure, and DevOps workflows. • Conducted Human-in-the-Loop (HITL) reviews by comparing model responses, identifying inaccuracies, and providing structured feedback for model improvement. • Designed prompt libraries and evaluation scenarios for code generation, debugging, refactoring, documentation generation, and code transformation tasks. • Identified hallucinations, syntax errors, security vulnerabilities, and inefficient implementations in generated outputs. • Collaborated with AI engineers and product teams to improve prompt engineering strategies, dataset quality, and model evaluation processes. • Evaluated outputs from multiple LLM providers, including OpenAI and Anthropic models, to assess response quality and consistency. • Supported Retrieval-Augmented Generation (RAG) initiatives by validating contextual responses generated using vector databases and knowledge sources. • Assisted in developing automated workflows for collecting, processing, and analyzing AI evaluation results.