Binance Accelerator Program Backend Engineer (AI Pro Agent Infrastructure)
Job Location:
Taipei City - Taiwan
Monthly Salary:
Posted on:
Vacancies:
Job Summary
About the Team
Binance AI Pro is building the next generation of AI-native experiences within Binance.
At the core of AI Pro is an agentic system that can understand user intent reason through multi-step tasks retrieve information call tools and safely execute actions across Binances ecosystem.
As a Backend Engineer Intern youll work directly on the Agent Runtime and AI infrastructure behind these experiences from intent routing and tool orchestration to evaluation observability and reliability. Were looking for strong technical students who enjoy going deep people who are excited byLLM Agents systems algorithms and experimentation and wants to work on real production AI systems alongside experienced Backend Data Science and Algorithm engineers.
You will have the opportunity to turn ideas fromprototype benchmark production and see your work directly impact how AI Pro behaves at scale.
Responsibilities
- Build and improve core components of the Agent Runtime including intent routing query rewriting RAG tool/skill orchestration and multi-step agent workflows.
- Design and optimize Tool Server / Tool Calling capabilities including tool discovery execution result handling and integration with web search market data and internal services.
- Develop evaluation datasets and evaluation harnesses to measure routing accuracy answer quality tool-use performance and agent reliability.
- Investigate and improve Agent quality including failure analysis regression detection prompt / workflow optimization and guardrail effectiveness.
- Improve system observability and reliability through tracing metrics logging dashboards and production monitoring.
- Debug issues across asynchronous services and agent pipelines and build integration tests using mocked LLM responses and replayable evaluation cases.
- Work closely with Backend Data Science and Algorithm engineers to experiment benchmark and bring AI capabilities into production.
Requirements (Must Have)
- Strong Python programming skills and solid software engineering fundamentals.
- Strong understanding of algorithms data structures and problem solving; able to reason about system behavior and trade-offs.
- Hands-on experience with LLM applications or Agent systems through research coursework internships or projects.
- Familiarity with concepts such as RAG Tool Calling Function Calling Prompting Agent workflows or LLM evaluation.
- Able to work comfortably with real codebases APIs asynchronous services testing CI/CD and code review.
- Strong debugging and analytical ability with a mindset of understand why it fails not just make it work.
Nice-to-have
- Experience with Agent frameworks such as LangGraph LangChain AgentScope LlamaIndex or similar.
- Experience with MCP / Tool ecosystems / Agent Runtime.
- Experience building Benchmark / Evaluation / LLM testing infrastructure.
- Familiarity with vector search embeddings RAG evaluation or LLM observability.
- Experience with Redis Kafka Docker Kubernetes or cloud-native systems.
- Exposure to LLM security guardrails prompt-injection defense or tool-use safety.
- Experience with model inference latency optimization or cost optimization.
What Youll Get
- Ship real features to a production AI Agent used by Binance users.
- Work on the Agent Runtime and AI infrastructure rather than only application-layer development.
- Learn from senior Backend Data Science and Algorithm engineers.
- Gain hands-on experience across the modern LLM application stack: Agent Runtime RAG Tool Calling Evaluation LLMOps and AI Safety.
- Own meaningful engineering projects end-to-end from problem definition implementation evaluation production.
Required Experience:
Unclear Seniority
About Company
Binance is a blockchain ecosystem comprised of Exchange, Labs, Launchpad, Info, Academy, Trust Wallet, and Blockchain Charity Foundation (BCF). Binance Exchange is one of the fastest and most popular cryptocurrency exchange platforms in the world, capable of processing over 1.4 millio ... View more
