代理基础设施工程师—核心Harness（Superagent）
Agent Infrastructure Engineer — Core Harness (Superagent)
ABOUT IMAGINEART
我们正在重新定义世界创造和设计的方式。
ImagineArt是全球增长最快的GenAI公司之一。我们的增长速度超过了大多数获得资金的初创公司——而且没有任何外部资金。
- 今年ARR超过3500万美元
- 社交媒体曝光量超过1亿次
- 自研图像生成模型，目前在照片真实度方面排名全球第三
没有融资，没有捷径。只有一支高效、有动力的团队正在打造全球最强的GenAI产品之一——而我们才刚刚开始。
我们正在寻找一名Agent Infrastructure Engineer来负责Superagent，这是我们核心的代理框架，支撑着我们AI产品中的对话、工具调用和多步骤代理工作流。
这是一个深入系统和基础设施的角色——不是提示工程，也不是简单地封装模型API。你将负责核心编排循环、工具调用基础设施、上下文和内存管理、流处理、重试、评估、可观测性以及性能优化。
KEY RESPONSIBILITIES
- 负责Superagent核心代理框架的架构、开发与演进
- 设计并优化代理执行循环，以降低延迟、提高可靠性、提升令牌效率、降低成本并提高任务完成率
- 构建和改进核心框架系统，包括上下文管理、内存/状态处理、工具路由、函数模式、结构化输出、重试和错误恢复
- 构建和维护代理评估基础设施，通过数据衡量质量并指导工程决策
- 集成和基准测试多个LLM供应商和模型，评估性能、成本、可靠性和能力
- 实现性能优化，如缓存、批处理、并行工具执行和提示/上下文压缩
- 在代理运行过程中构建深度可观测性和仪器仪表，包括追踪、日志、指标和回归检测
- 当现有抽象不足以满足需求时，扩展和定制底层代理框架
- 构建与不断发展的AI和工具生态系统的可靠集成
- 与产品工程团队紧密合作，在平台背后保持框架复杂性的同时，提供清晰的抽象接口
- 调试并解决非确定性、分布式和模型驱动系统中的复杂问题
REQUIRED SKILLS & QUALIFICATIONS
- 4年以上软件工程、后端工程或系统基础设施经验
- 精通Python和/或TypeScript
- 有实际操作经验
查看英文原文
ABOUT IMAGINEART
We're redefining how the world creates and designs.
ImagineArt is one of the fastest-growing GenAI companies in the world. We've scaled faster than most funded startups — with zero outside funding.
- $35M+ ARR crossed this year
- 100M+ social impressions
- Built and shipped our own image generation model, now ranked #3 globally for photo realism
No funding. No shortcuts. Just a sharp, driven team building one of the strongest GenAI products in the world — and we're just getting started.
We're looking for an Agent Infrastructure Engineer to own Superagent, our core agent harness that powers conversations, tool calls, and multi-step agentic workflows across our AI products.
This is a deep systems and infrastructure role — not prompt engineering and not simply wrapping model APIs. You'll work on the core orchestration loop, tool-calling infrastructure, context and memory management, streaming, retries, evaluation, observability, and performance.
KEY RESPONSIBILITIES
- Own the architecture, development, and evolution of Superagent, our core agent harness.
- Design and optimize the agent execution loop for latency, reliability, token efficiency, cost, and task completion.
- Build and improve core harness systems including context management, memory/state handling, tool routing, function schemas, structured outputs, retries, and error recovery.
- Build and maintain agent evaluation infrastructure to measure quality and guide engineering decisions with data.
- Integrate and benchmark multiple LLM providers and models, evaluating performance, cost, reliability, and capabilities.
- Implement performance optimizations such as caching, batching, parallel tool execution, and prompt/context compression.
- Build deep observability and instrumentation across agent runs, including tracing, logging, metrics, and regression detection.
- Extend and customize underlying agent frameworks when existing abstractions are insufficient.
- Build reliable integrations with evolving AI and tool ecosystems.
- Work closely with product engineering teams to expose clean abstractions while keeping harness complexity behind the platform.
- Debug and resolve complex issues across non-deterministic, distributed, and model-driven systems.
REQUIRED SKILLS & QUALIFICATIONS
- 4+ years of experience in software engineering, backend engineering, or systems infrastructure.
- Strong proficiency in Python and/or TypeScript.
- Hands-on experience building or operating LLM-based agents in production.
- Strong understanding of tool calling, function schemas, context limits, structured outputs, model failures, and unreliable LLM behavior.
- Experience with at least one agent framework such as LangGraph, OpenAI Agents SDK, CrewAI, AutoGen, or a custom/homegrown agent harness.
- Strong understanding of agent orchestration and multi-step workflows.
- Experience building or working with evaluation suites, benchmarks, A/B testing, or other measurement systems for AI products.
- Strong understanding of concurrency, caching, profiling, performance optimization, and latency/cost tradeoffs.
- Experience working with LLM APIs and production AI infrastructure.
- Excellent debugging and problem-solving skills, especially for complex and non-deterministic systems.
- Passionate about technology, self-driven, and proactive with a strong builder mindset.
OPTIONAL / NICE-TO-HAVE SKILLS
- Contributions to open-source agent frameworks, LLM tooling, or AI infrastructure.
- Experience with RAG pipelines, vector databases, or long-term memory systems for AI agents.
- Familiarity with MCP (Model Context Protocol) or similar tool-integration standards.
- Experience with LLM inference infrastructure, model routing, rate limits, fallbacks, or high-volume model APIs.
- Experience with LangChain, LlamaIndex, LangGraph, DSPy, or similar AI infrastructure frameworks.
- Experience with Kubernetes, Docker, cloud infrastructure, or distributed systems.
- Experience building internal developer platforms or infrastructure used by multiple engineering/product teams.
- Strong background in observability, distributed tracing, and production reliability.
- Contributions to open-source projects or personal AI infrastructure projects.
WHY JOIN US?
- Own the core agent infrastructure behind our AI products — every improvement you make can multiply across the entire platform.
- Work on real production-scale AI systems, not demo agents or simple API wrappers.
- Solve challenging problems across LLMs, distributed systems, orchestration, performance, and infrastructure.
- Have direct influence over the architecture and technical roadmap of our entire agent stack.
- Collaborate with a passionate and talented team building some of the most ambitious GenAI products in the market.
- Competitive salary and benefits package.
- A culture that encourages ownership, experimentation, learning, and data-driven engineering.