# Job: 字节 Seed - 大模型 Agent 研究工程师

---
type: job
company: 字节 Seed
role: 大模型Agent研究工程师-Seed
location: unknown
source_url: https://jobs.bytedance.com/experienced/position/7628902314323806469/detail
source_name: 字节跳动招聘
source_quality: official
collected_at: 2026-07-08
posted_at:
first_seen_at: 2026-07-08
last_checked_at: 2026-07-08
snapshot_type: new
job_code: 7628902314323806469
previous_snapshot:
status: unknown
level: experienced
team: Seed
business_area: agent research and engineering
employment_type: full-time
salary:
skills:
  - agent
  - memory
  - context-engineering
  - planning
  - multi-agent
  - llm-application
topics:
  - harness
  - memory
  - context-compression
models:
  - Seed
related_papers:
  - 2026-memory-agent-survey
  - 2026-evomembench
related_experiments:
  -
related_projects:
  -
relevance: high
---

## Source Snapshot

- source: 字节跳动官方招聘页；网页正文由前端渲染，当前采集依赖搜索摘要。
- collected_at: 2026-07-08
- availability: URL 可定位到官方岗位页，但需要后续用浏览器或 API 抽取完整 JD。

## Raw JD Summary

搜索摘要显示该岗位负责大模型 Agent 系统研究与工程落地，围绕复杂任务执行持续优化 Agent 效果，关键词包括 harness、memory、context compression、planning/execution、multi-agent。

## Extracted Signals

- hard requirements: agent harness, memory, context compression, planning/execution, multi-agent。
- hidden signals: Seed 把 Agent 作为研究与工程落地结合的问题，而不是纯应用开发。
- business direction: 复杂任务执行和通用 Agent 能力。

## Knowledge Links

- concepts: context engineering, agent harness, memory
- papers: Memory survey, EvoMemBench
- experiments: context compression comparison
- projects:

## Gaps for Us

- 需要后续用可重复脚本抓取字节招聘 JS 数据或手动浏览器快照。