AI Software Engineer – Morningtrain
Designed and generated over 1,200 multi-turn conversations for training large language model (LLM) agents. Created evaluation datasets and engineered dialogue flows using structured JSON schemas to benchmark reasoning and improve dataset diversity. Refined annotation playbooks to improve dataset acceptance rate and reduce invalid formatting in internal QA reviews. • Simulated user and assistant dialogues across 6–12 conversational turns • Developed function-calling scenarios for tool use and task automation • Focused on realistic intent, context tracking, and edge case coverage • Enhanced internal QA through iterative review and scenario refinement