Agent策略优化工程师 - 火山引擎
Core
Design and optimize core Agent loops (Planning, Reasoning, Tool Use, Reflection) to improve success rates, stability, and efficiency for complex long-horizon tasks.
Role type
Senior IC Agent Strategy Optimization Engineer
Builds
Autonomous agents capable of executing complex, multi-step tasks with context management and error recovery
Domain
Cloud AI / Generative AI / Agentic Systems
Deliverable
production ML models
Required skills
Large language model (LLM) architecture, Prompt Engineering, RAG, Context Engineering, Python, Agent frameworks (LangGraph, LlamaIndex, AutoGen, Dify, CrewAI), ReAct, Reflexion, Plan-and-Execute, Tree of Thoughts, SFT, DPO, Agentic RL (GRPO, DAPO), Agent evaluation frameworks (SWE-bench, GAIA, AgentBench)
Preferred skills
Experience with long-horizon execution, data generation for SFT/RL, building data loops from online traces
Technologies
Transformer, GPT, Claude, Gemini, Qwen, DeepSeek, MCP
Responsibilities
Design and optimize Agent core linkages for complex task execution; Develop context engineering and memory systems (short/long-term, semantic); Construct data engines for high-quality SFT/RL data generation; Implement post-training strategies (SFT/DPO/Agentic RL) for continuous evolution; Build multi-level evaluation systems (offline/online A/B/case regression)
Seniority
Senior, hands-on IC
