2 months ago
Taipei, Taiwan or Singapore, SingaporeEntry Level
Responsibilities
- Design and operate adaptive, self-correcting, multi-hop retrieval pipelines and Agentic RAG systems.
- Implement agent capabilities including context management, long-term memory, subagents, multi-agent architectures, self-evolving agents, and real-world task execution.
- Define and build benchmarks, datasets, annotation strategies, and evaluation methods for agent and RAG performance.
- Use user feedback and real-world production task data to design experiments and improve agent and retrieval performance.
- Develop agent harness features such as session recovery, sandbox isolation, middleware and hook systems, multi-tenant runtimes, plan/execute loops, and retrieval-grounded tool calling.
- Rapidly prototype and evaluate original ideas through iterative research and engineering workflows.
Requirements
- At least 1 year of hands-on production experience with LLM, RAG, and AI agent systems.
- Hands-on experience building end-to-end retrieval pipelines using embedding models, vector stores, hybrid search, reranking, chunking, text cleaning, and multimodal data parsing.
- Experience implementing Agentic RAG patterns such as Self-RAG, Corrective RAG, adaptive retrieval, multi-hop decomposition, and retrieve-reflect-refine loops.
- Experience with agent harness runtimes such as Pi Agent, AgentScope 2.0, or equivalent orchestration frameworks.
- Deep familiarity with LLM APIs, KV Cache, agent loops, tool use, reasoning, planning, skills, MCP, memory, subagents, and multi-agent systems.
- Strong prompt engineering and context engineering knowledge, independent research ability, and capacity to translate ideas into runnable prototypes.
- Proficiency with AI-assisted software development workflows and the ability to work across unfamiliar languages, frameworks, and domains.
- Preferred experience with Claude Code, OpenClaw, Cowork, Manus, RAGAS, TruLens, GraphRAG, knowledge graphs, model training, RLHF, LiteLLM, Kubernetes/EKS, prompt-injection defense, sandbox hardening, or guardrail design.
Benefits
- Work-from-home arrangement, with the specific arrangement varying according to business-team work requirements.
- Opportunity to work on Binance’s global blockchain ecosystem and collaborate with world-class talent.
- Fast-paced, autonomous projects in a flat, user-centric organization.
- Career growth and continuous learning opportunities.
- Company benefits and competitive salary.
Tech Stack
Categories
About Binance
Binance builds a global cryptocurrency exchange and related infrastructure for retail and institutional traders, developers, and businesses. Its products include spot and derivatives trading, fiat on-ramps, wallets, payments, staking, custody, and Web3/NFT services, with revenue driven largely by trading and transaction fees. Founded in 2017 and privately held, it operates worldwide and also stewards the BNB Chain ecosystem, with its main exchange regularly ranking first by crypto trading volume.