对话式AI工程师 - 游戏原型
Conversational AI Engineer - Game Prototype
远程 · 至少与美国太平洋时间有4小时重叠 · 合同工 · 初期3个月合作
职位描述
我们正在开发一款新的游戏，其中 AI 驱动的 NPC 对话是玩家体验的核心部分。
我们正在寻找一名中级对话 AI 工程师，帮助我们设计和构建这些互动背后的 AI 技术栈。
核心问题：玩家与 NPC 对话，NPC 以角色身份大声回答，速度快到足以让人感觉像是一场对话——目标是低于1秒。如果你曾经构建过这样的循环并测量过时间消耗，我们希望听到你的消息。如果没有，这个职位不适合你。
这是一个动手实践的工程职位。你将选择技术，设计对话架构，连接各个部分，并在 Unity 中构建系统。
你将负责原型对话 AI 系统的技术实现，包括：
- 设计并构建玩家与 NPC 之间的端到端语音对话流程。
- 评估并集成语音转文字（STT）、大语言模型（LLM）、文字转语音（TTS）和语音转语音技术。
- 评估不同的对话记忆、上下文、个性和 NPC 行为方法。
- 原型实时对话交互并优化延迟、质量、可靠性和成本。
- 确定哪些组件应在设备上运行，哪些应在云端运行。
- 在 Unity/C#中构建并集成系统。
- 建立一个可以从原型演进为生产系统的架构。
你将与游戏设计和工程团队紧密合作，确定技术需要完成什么以及如何让它对玩家来说显得自然且响应迅速。
我们的技术栈
我们没有固定的技术栈。你将选择它，并用你测量的数字来证明每个选择：响应时间、每分钟对话成本以及失败频率。
我们寻找的人选
必须具备（全部三点）：
- 5年以上专业软件工程经验。
- 至少有一个使用 Unity 和 C#开发的已发布项目（一款已发布的游戏、应用或生产工具）。你将被要求提供链接。
- 自己构建的完整语音对话循环，从语音输入到语言模型再到语音输出。你将被要求提供链接或演示。
加分项：
- 游戏 AI、NPC 系统或互动娱乐
- 对话记忆和代理架构
- 设备端语音或语言模型
- 多语言或语言导向的 AI 系统
不适合的人选
- 你
查看英文原文
Remote · at least 4 hours overlap with US Pacific · Contractor · Initial 3-month engagement
The Role
We are prototyping a new game where AI-driven conversations with NPCs are a core part of the player experience.
We're looking for a mid-level Conversational AI Engineer to help us design and build the AI technology stack behind those interactions.
The core problem: a player speaks to an NPC, and the NPC answers out loud, in character, fast enough to feel like a conversation - target, e.g. under 1 second. If you have built a loop like that and measured where the time goes, we want to hear from you. If you haven't, this role is not a fit.
This is a hands-on engineering role. You will choose the technologies, design the conversational architecture, connect the pieces and build the system in Unity.
What You'll Do
You will own the technical implementation of the prototype's conversational AI system, including:
- Design and build the end-to-end voice conversation pipeline between players and NPCs.
- Evaluate and integrate speech-to-text (STT), LLM, text-to-speech (TTS), and speech-to-speech technologies.
- Evaluate different approaches to conversational memory, context, personality, and NPC behavior.
- Prototype real-time conversational interactions and optimize for latency, quality, reliability, and cost.
- Determine which components should run on-device versus in the cloud.
- Build and integrate the resulting systems into Unity/C#.
- Establish an architecture that can evolve from a prototype into a production system.
You'll work closely with the game design and engineering team to determine what the technology needs to accomplish and how best to make it feel natural and responsive to players.
Our Stack
We have no fixed stack. You will choose it, and you will justify each choice with numbers you measured: response time, cost per conversation minute and how often it fails.
What We're Looking For
Must have (all three):
- 5+ years of professional software engineering.
- At least one shipped project built in Unity with C# (a released game, app or production tool). You will be asked to link it.
- A working voice conversation loop you built yourself, end to end: speech in, language model, speech out. You will be asked to link or demo it.
Nice to have:
- Game AI, NPC systems or interactive entertainment
- Conversational memory and agent architectures
- On-device speech or language models
- Multilingual or language-focused AI systems
Not A Fit If
- Your AI experience is mostly prompt writing, or chat apps built on a single API.
- You have not shipped anything in Unity.
- You want to research or train models rather than build a playable prototype.
Who You'll Be
You'll be deciding what sits between the player's voice and the NPC's reply, not implementing an architecture someone hands you.
You like getting from "let's try this" to a working build fast, with little process and a lot of ownership.
Engagement
Remote · at least 4 hours overlap with US Pacific · Contractor · Initial 3-month engagement
The initial engagement will focus on designing and building a working prototype of the conversational AI system in Unity.
For the right person, there may be an opportunity to continue working with the team as the project moves beyond the prototype stage.
To Apply
We look at the work before the resume. Applications without all three items below are not reviewed.
- Your resume and cover letter.
- A link to a voice or conversational system you built (repo, video, live demo or write-up). Tell us which parts you built yourself.
- Short answers, a few sentences each, in your own words:
- In that system, what was the time from the player finishing speaking to the reply starting, and where did that time go?
- What did you run on-device and what in the cloud, and why?
- What would you try first to make an NPC remember a player between sessions?
Shortlisted people will walk us through their demo on a 30-minute call.
Originally posted on Himalayas