At RWS AI Labs, I evaluate and rank 300+ LLM responses per week for accuracy, helpfulness, safety, and instruction-following. I review multi-turn dialogues for hallucinations and factual errors, maintaining 96% agreement with client QA rubrics on Labelbox.
At magnitron Data Systems, I built and trained a Python NLP text classifier with PyTorch that reached 91% F1 on a 50,000-record support-ticket dataset. As a freelance Web Developer, I designed and delivered eight responsive websites using React and Node.js, cutting average page load time by 40%.

