AI Agent Engineer
Binance is a prominent force in the global blockchain space, known for operating the largest cryptocurrency exchange worldwide. Trusted by millions across numerous countries, Binance excels in areas such as security, transparency, trading speed, liquidity, and a diverse range of digital products. The company's offerings span trading, finance, education, research, institutional services, and more, utilizing digital assets and blockchain to create an all-inclusive financial ecosystem that promotes financial freedom and enhances global financial accessibility.
We are seeking a research-oriented engineer to join our AI Infra team, focusing on the convergence of cutting-edge model capabilities and real-world agent deployment. In this role, you will collaborate closely with researchers and engineers to expand the boundaries of AI agent functionalities, covering aspects like Agentic RAG systems, context management, task execution, self-evolving agents, and multi-agent coordination.
This is a hybrid role that combines elements of both engineering and research. The ideal candidate will be innovative, capable of conceptualizing original ideas, conducting experiments, developing prototypes, and iterating quickly based on real-world feedback. We are looking for someone who integrates agent tools seamlessly into their daily workflow and possesses strong insights into model behavior.
Responsibilities:
- Design and operationalize advanced retrieval pipelines, transitioning from static patterns to adaptive, multi-hop workflows with self-correcting mechanisms and dynamic retrieval controls.
- Work closely with researchers and engineers to implement model-driven innovations, emphasizing context management, long-term memory utilization, architectural frameworks for subagent and multi-agent systems, self-evolving agents, and real-world task execution.
- Develop and introduce domain-specific and retrieval-focused benchmarking strategies and evaluation protocols to enhance agent intelligence, focusing on metrics such as retrieval efficiency, groundedness, latency, and task success rates.
- Foster real-world feedback mechanisms by utilizing user data and task-related feedback loops, designing experiments and datasets to continually refine and optimize agent and retrieval performance for production scenarios.
Requirements:
- Demonstrated hands-on experience (1+ year) with LLM, RAG, and AI agent systems within operational environments.
- Proficiency in RAG and Agentic RAG engineering, encompassing the establishment of end-to-end production retrieval pipelines, handling model embeddings, vector stores, hybrid search strategies, and adaptive retrieval techniques.
- Practical experience with Agent Harness runtimes like Pi Agent, AgentScope 2.0, displaying skills in session recovery, sandbox isolation, middleware systems, and multi-tenant runtime operations.
- In-depth understanding of LLM and agent fundamentals, including APIs, reasoning mechanisms, memory management, subagent architectures, and multi-agent collaboration, as well as strong knowledge of prompt and context engineering.
- Proven capability in independent research, from analyzing complex problems to generating novel ideas and transforming them into tangible prototypes through rapid iterations based on empirical evidence.
- Proficient in AI-native engineering, displaying adept skills in rapid development using AI-assisted workflows across diverse languages, frameworks, and domains.
Nice to Have:
- Extensive hands-on experience with agent products such as Claude Code, OpenClaw, Cowork, or Manus, integrated into personal workflow.
- Familiarity with RAG evaluation tools like RAGAS or TruLens, specialized in benchmarking retrieval quality, groundedness, and latency profiling.
- Expertise in GraphRAG or knowledge graph-augmented retrieval techniques.
- Proficiency in Pi Agent, AgentScope 2.0, or similar Agent Harness platforms, showcasing skills in middleware composition, session management, and sandbox backends.
- Background in model training, RLHF, or model-system co-design, as well as experience with LiteLLM or multi-provider proxies.
- Knowledge of Kubernetes/EKS practices such as pod isolation, resource management, and secrets handling, along with expertise in security engineering aspects like prompt injection defense and sandbox fortification.
Discover the potential of shaping a bright future within the blockchain ecosystem with Binance, where you can collaborate with top-tier professionals in an innovative, user-centric environment. Benefit from autonomy in working on challenging and unique projects, within a results-oriented setting that fosters continuous learning and career growth opportunities. Enjoy competitive compensation and company benefits, coupled with a flexible work arrangement that may include remote working options tailored to specific team requirements.imensional-environment.
