AI Agent Architect

About Us

岗位概述

核心职责

1. Agent 体系架构设计：主导设计 AI 智能体架构，定义 Agent 在 Openclaw 平台上的运行模型、生命周期管理与资源隔离策略；构建支撑交易辅助、风控研判、合规监察等金融核心场景的 Agent 能力基座，确保系统在高并发、低延迟、强合规约束下稳定运行。
2. 金融场景 Agent 工程化落地：将交易信号分析、风险识别、监管合规、智能客服等业务需求转化为可执行的 Agent 工作流；设计任务规划链路（ReAct / Plan-and-Execute）、异常回退机制与结果验证体系，推动 Agent 从 PoC 快速进入生产环境。
3. Openclaw 平台建设与演进：深度参与 Openclaw Agent Runtime 的架构演进与性能优化；设计插件体系、能力注册中心与多 Agent 编排机制，降低业务团队接入成本，持续提升平台的扩展性与复用率。
4. 工具生态与内部系统集成：构建标准化工具调用框架，打通 Bybit 内部交易引擎、风控系统、数据平台与外部 API；设计工具权限管控、调用链追踪与错误恢复机制，集成 MCP 等标准协议，构建金融级异构工具生态。
5. 生产质量与可观测性体系：主导 Agent 系统的可观测性建设，包括 Trace 追踪、Evaluation 评测、成本监控与延迟治理；建立金融场景下的 Agent 质量基线与回归测试体系，持续提升系统稳定性与可解释性。
6. 技术前瞻与团队赋能：持续追踪 Agent 领域前沿进展，将研究成果转化为 AI 工程实践；制定团队 Agent 开发规范与最佳实践，通过 Code Review、技术分享与方向评审推动团队整体 AI-native 工程能力提升。

任职要求

1. 扎实的 LLM/Agent 工程经验：具备 3年以上 AI 系统开发经验，其中 1年以上 大型语言模型（LLM）、RAG 或 Agent 系统的生产环境落地经验；具备 Openclaw 平台的开发与优化实战经验，熟悉其 Agent Runtime 架构、插件体系与多 Agent 编排机制，能基于 Openclaw 独立完成复杂 Agent 应用的工程化落地与性能调优。
2. 系统架构能力：具备分布式系统设计能力，熟悉微服务架构、消息队列（Kafka / RabbitMQ）、异步任务调度等工程模式；能够独立完成从需求分析、架构设计到技术选型的完整决策链路。
3. 编程能力：熟悉至少一门后端语言（Golang / Java / Node.js）；具备 REST API 设计与 Docker / Kubernetes 容器化部署能力。
4. LLM 核心技术理解：深入理解 Transformer 架构、Attention 机制、Prompt Engineering（CoT、Few-shot、Structured Output）；熟悉主流模型 API（OpenAI、Anthropic Claude、Google Gemini）及本地部署方案（Ollama、vLLM）。
5. 工程质量意识：具备良好的代码设计习惯（设计模式、单元测试、CI/CD）；有构建 LLM 评测（Evaluation）体系或 Agent Tracing（LangSmith / Phoenix / Langfuse）的实战经验。

加分项

• 有 MCTS、A\*、强化学习（RL） 与 LLM 结合的规划/推理研究或工程经验，熟悉 SFT / RLHF 训练流程。
• 在 金融、Web3、合规 或其他高复杂度垂直领域有 Agent 系统落地经验，理解领域知识对 Agent 设计的特殊约束。
• 有开源贡献或在 NeurIPS / ICML / ACL / ICLR 等顶会发表过 Agent / RAG / LLM 相关论文。
• 具备 多模态（Multimodal） Agent 开发经验，包括图文理解、文档解析（PDF / 财报 / 研报）等非结构化数据处理能力。
• 熟悉 MCP（Model Context Protocol） 或参与过 Agent 标准协议的设计与实现。

About the Role

Key Responsibilities

1. Agent Architecture Design: Lead the architectural design of AI's intelligent Agent layer — defining agent execution models, lifecycle management, and resource isolation strategies on the Openclaw platform. Build the foundational Agent capabilities powering core financial scenarios including trading intelligence, risk assessment, and compliance monitoring, engineered for high-concurrency, low-latency, and regulatory-grade reliability.
2. Financial Scenario Agent Engineering: Translate complex financial use cases — trading signal analysis, risk identification, regulatory compliance, and intelligent customer service — into production-ready Agent workflows. Design task planning chains (ReAct / Plan-and-Execute), fallback and error recovery mechanisms, and result validation frameworks that accelerate the path from PoC to production.
3. Openclaw Platform Development & Evolution: Contribute deeply to the architectural evolution and performance optimization of the Openclaw Agent Runtime. Design the plugin ecosystem, capability registry, and multi-agent orchestration mechanisms to reduce onboarding friction for product teams and improve platform extensibility and reusability.
4. Tool Ecosystem & Internal System Integration: Build a standardized tool-calling framework connecting Bybit's internal trading engine, risk systems, data platforms, and external APIs. Design tool permission controls, call-chain tracing, and error recovery; integrate standards such as MCP to unify a financial-grade heterogeneous tool ecosystem.
5. Production Quality & Observability: Lead the observability stack for Agent systems, including trace monitoring, evaluation pipelines, cost tracking, and latency governance. Establish quality baselines and regression testing frameworks for financial Agent scenarios, continuously improving system stability and interpretability.
6. Technical Vision & Team Enablement: Track and internalize cutting-edge Agent research, translating advances into AI engineering practice. Define team-wide Agent development standards and best practices; drive AI-native engineering capability uplift through code reviews, technical sharing, and architectural reviews.

Requirements

1. Proven LLM / Agent Engineering Experience: 3+ years of AI systems development experience, including 1+ year of production experience with Large Language Models (LLMs), RAG systems, or Agent frameworks; hands-on development and optimization experience with Openclaw, including familiarity with its Agent Runtime architecture, plugin ecosystem, and multi-agent orchestration mechanisms.
2. Systems Architecture Proficiency: Strong distributed systems design skills including microservices architecture, message queues (Kafka / RabbitMQ), and async task scheduling; capable of independently driving the full decision chain from requirements analysis through architecture design to technology selection.
3. Programming Expertise: Proficiency in at least one backend language (Golang / Java / Node.js); solid REST API design and containerized deployment skills (Docker / Kubernetes).
4. Deep LLM Technical Understanding: Strong grasp of Transformer architecture, Attention mechanisms, and advanced Prompt Engineering (CoT, Few-shot, Structured Output); experience with major model APIs (OpenAI, Anthropic Claude, Google Gemini) and local deployment solutions (Ollama, vLLM).
5. Engineering Quality Standards: Strong software design fundamentals (design patterns, unit testing, CI/CD); hands-on experience building LLM evaluation frameworks or Agent tracing pipelines (LangSmith / Phoenix / Langfuse).

Nice to Have

• Experience combining MCTS, A\*, or Reinforcement Learning (RL) with LLMs for planning and reasoning; familiarity with SFT / RLHF training pipelines.
• Proven Agent system deployments in high-complexity vertical domains such as Finance, Web3, or Compliance, with an understanding of domain-specific constraints on Agent design.
• Open-source contributions or publications in NeurIPS / ICML / ACL / ICLR on Agent, RAG, or LLM-related topics.
• Experience with multimodal Agent development, including vision-language understanding and unstructured document parsing (PDFs, financial reports, research papers).
• Familiarity with MCP (Model Context Protocol) or prior involvement in designing and implementing Agent communication standards.

Why Join Us
At Bybit, we are committed to fostering a supportive and enriching work environment. 
Our benefits include:
- Study Growth Fund: We support your professional development and continuous learning.
- Internal Events: Participate in regular team-building activities, workshops, and events designed to promote collaboration and innovation.
- Global Collaboration: Be part of a diverse, international team, working alongside colleagues from around the world.
- Career Advancement: Access opportunities for growth and advancement within a rapidly expanding global company.
- Internal Mobility: Grow with us- Your long-term development is important to us. We offer internal job opportunities to help build your career path.