LLM Conversation Trainer- South Korea
Budget: $8 – $15 USD
I’m putting together a small team to craft high-quality, multi-turn Korean conversations that will train a next-generation large language model. Your task is to write realistic back-and-forth exchanges between a user and an AI assistant, all in ko_KR, and to show the assistant working through tasks that involve email—drafting, replying, summarising, scheduling sends, and so on.
I’ve already designed clear guidelines, JSON schemas and API reference notes; you’ll simply plug your dialogues into those formats. Because the material lives inside a structured dataset, each turn must feel natural and helpful while also staying logically consistent and technically sound. Advanced written English is essential because all meta-comments, variable names and reviewer feedback are in English.
You’ll be collaborating with other contributors, reviewing one another’s work, and refining prompts so a reviewer can approve them without major edits. I expect you to bring solid reasoning skills, a knack for problem-solving and at least some previous hands-on experience with AI or LLM projects. Familiarity with email workflows is mandatory; if you also know how to simulate calendars or maps, that’s a bonus but not required.
Deliverables (per assignment)
• A batch of multi-turn dialogues (exact count shared upon kickoff) in validated JSON format
• A short self-check report explaining any edge-case decisions
I review submissions within 24 hours and give clear, actionable feedback. Once a batch meets the acceptance criteria in the style guide, I release the milestone and open the next one.
If you’re ready to help shape how Korean users will interact with AI, let’s get started.
I’ve already designed clear guidelines, JSON schemas and API reference notes; you’ll simply plug your dialogues into those formats. Because the material lives inside a structured dataset, each turn must feel natural and helpful while also staying logically consistent and technically sound. Advanced written English is essential because all meta-comments, variable names and reviewer feedback are in English.
You’ll be collaborating with other contributors, reviewing one another’s work, and refining prompts so a reviewer can approve them without major edits. I expect you to bring solid reasoning skills, a knack for problem-solving and at least some previous hands-on experience with AI or LLM projects. Familiarity with email workflows is mandatory; if you also know how to simulate calendars or maps, that’s a bonus but not required.
Deliverables (per assignment)
• A batch of multi-turn dialogues (exact count shared upon kickoff) in validated JSON format
• A short self-check report explaining any edge-case decisions
I review submissions within 24 hours and give clear, actionable feedback. Once a batch meets the acceptance criteria in the style guide, I release the milestone and open the next one.
If you’re ready to help shape how Korean users will interact with AI, let’s get started.