Agent Innovative Application Algorithm Engineer Agent创新应用算法工程师

职位要求 / 描述

About us Bitget is one of the world's leading digital assets ecosystems. With over 120 million registered users, Bitget has one of the most comprehensive suites of blockchain products and services available via bitget.com. Our mission is to support the growth of the digital assets industry and we believe it represents the future of finance. What we do empowers the future of finance by ensuring secure, efficient and smart digital transactions. We are one of the fastest growing companies in the digital asset sector. If you are looking for cutting-edge work, where you will have opportunities to develop your career among peers who are experts in their field, and you believe in the future of digital currency, then look no further than Bitget! What you'll do 1. Leading the exploration and incubation of innovative products for Agent technology Actively keep track of the latest advancements in the field of AI Agents - including but not limited to the MCP/A2A interoperability protocol, Deep Research Agent, the Agentic Coding paradigm (such as the design philosophy of Anthropic Claude Code), the Skills modular architecture (such as the Skills model of OpenClaw/OpenHands), multi-Agent collaboration systems, and Agent driven by reasoning models - combined with the business scenarios in the virtual currency domain, independently initiate the conception, feasibility verification, and prototype construction of innovative products. We expect you to be the one who actively defines the problem, rather than waiting for the input of requirements. 2. Responsible for the algorithm design and large-scale implementation of the Agent system Design and implement the Agent algorithm solutions for scenarios such as analysis of virtual currency on-chain assets, transaction decision assistance, in-depth research, and information aggregation. The core modules covered include: • Task Planning and Deep Reasoning (Planning & Extended Thinking): Following the design philosophy of "first deep thinking then action" in Anthropic Agentic Coding, combined with the adaptive search planning paradigm of Deep Research Agent, the Agent can autonomously break down complex financial tasks, conduct thorough reasoning, and execute multiple steps; • Skill Abstraction and Composable Architecture (Skills Composition): Drawing on the Skills model of OpenClaw, the capabilities of querying chain data, K-line technical analysis, contract risk auditing, public opinion monitoring, and transaction strategy execution are encapsulated into standardized and reusable Skill modules, enabling the Agent to dynamically select and orchestrate Skills based on tasks and continuously accumulate new Skills to achieve capability evolution; • Tool Invocation Orchestration (Tool-use Orchestration): Based on standard protocols such as MCP, realize standardized connections between the Agent and chain data sources, market interfaces, DeFi protocols, and analysis tools; • Multi-Agent Collaboration (Multi-Agent Collaboration): Drawing on the idea of A2A protocol, build a collaborative workflow for multiple Agents, supporting the dynamic orchestration and communication of roles such as research Agent, trading Agent, and risk control Agent; • Iterative Self-Verification and Correction (Iterative Self-Refinement): Deeply drawing on the "execution → verification → correction" closed-loop mechanism of Anthropic Claude Code, build the Agent's autonomous verification and iterative optimization capabilities in financial decision-making scenarios to ensure the reliability and accuracy of the output; • Autonomous Exploration of the Environment and Context Construction: Referencing the Agent's autonomous exploration of code repositories in Agentic Coding, enable the Agent to actively perceive and comprehensively understand the chain ecosystem and market environment, reducing reliance on manual information input; • Long-term Memory and Knowledge Accumulation: Support the Agent's continuous tracking, expe

技能关键字

职责方向