AI Training Engineer
TextRLHF
Fine-tuned and evaluated large language models (LLMs) using RLHF and instruction-tuning techniques to improve response quality. Developed and executed robust evaluation frameworks including automated benchmarks, red-teaming, and human-in-the-loop scoring. Collaborated with data scientists and product teams to ensure AI outputs align with safety guidelines and KPIs.
• Led RLHF reward modeling to reduce model hallucination rate by 27%
• Developed Python tools for dataset curation and annotation quality checks
• Integrated automated and human evaluation pipelines for safety and accuracy
• Improved factual accuracy through adversarial prompt engineering and scoring
Fine-tuned and evaluated large language models (LLMs) using RLHF and instruction-tuning techniques to improve response quality. Developed and executed robust evaluation frameworks including automated benchmarks, red-teaming, and human-in-the-loop scoring. Collaborated with data scientists and product teams to ensure AI outputs align with safety guidelines and KPIs. • Led RLHF reward modeling to reduce model hallucination rate by 27% • Developed Python tools for dataset curation and annotation quality checks • Integrated automated and human evaluation pipelines for safety and accuracy • Improved factual accuracy through adversarial prompt engineering and scoring
2025 - Present