Product Manager, AI Agent & Harness
On-siteHong Kong, Hong Kong
Job Summary
Own the design of Agent-related capabilities across Binance Agentic, including task flows, user interaction patterns, capability boundaries, failure fallbacks, and result feedback loops. Build and maintain the Harness framework by designing test scenarios, defining evaluation criteria, curating task sets, analyzing performance, and driving actionable iteration recommendations. Partner with Tech Owners, frontend engineers, and Skill design leads to break down business requirements into shippable, testable product specs. Map high-value, deployable Agent use cases across Web3 Wallet, Account, Payment, Trading, and other Binance business lines. Stay current on AI Agent, MCP, tool-calling, prompt engineering, and automated evaluation developments to translate them into executable product initiatives. Define and track Agent performance metrics, including task success rate, user satisfaction, error rate, response quality, and tool-calling accuracy. Proactively identify issues from real-world Agent usage and drive improvements across product, engineering, and operations.
Required Qualifications
- Experience in AI products, Agent products, AIGC applications, automation tools, or developer tooling
- Background in Web3, exchanges, wallets, payments, on-chain transactions, DeFi, or crypto assets
- Hands-on understanding of how AI Agents work — tool-calling, task planning, context management, and evaluation frameworks
- Strong ownership and learning agility; able to define problems, decompose workstreams, and drive execution in ambiguous environments
- Solid product fundamentals — clear PRDs, flow diagrams, test cases, acceptance criteria, and iteration plans
- Excellent cross-team collaboration and communication skills across engineering, design, operations, and business stakeholders
Desired Qualifications
- Experience with exchanges, Web3 wallets, on-chain trading, payments, or fintech products
- Experience with AI Agent Harness, Eval, Benchmarking, automated testing platforms, or prompt evaluation
- Familiarity with MCP, Function Calling, RAG, multi-Agent orchestration, or workflow automation
- Knowledge of the Binance ecosystem or mainstream Web3 user scenarios
- Strong English reading and writing skills, with the ability to keep up with global AI / Web3 product and technical materials
Hiring someone like this?
Get your role in front of qualified candidates on Sorce.