CareerPlanSign in

微信小微-大模型算法研究员-Harness合版与记忆方向

Beijing, China💼 Full-time🗓 2026-09-28

Core

Design and develop the Agent Harness framework for WeChat Assistant XiaoWei, focusing on agent scheduling, task orchestration, and memory systems.

Role type

Senior IC large model algorithm researcher (Agent Harness & Memory)

Builds

Production-ready Agent Harness framework and AI Agent memory systems for WeChat Assistant

Domain

Consumer social media / Large Language Models / AI Agents

Deliverable

production ML models

Required skills

Large model application development, Agent scheduling and task orchestration, RAG and memory retrieval, Post-training and model merging, Long context processing, System design

Preferred skills

vibe coding, Cross-session continuous memory, Edge-cloud collaborative architecture

Technologies

PyTorch, transformers

Responsibilities

Design and develop the Agent Harness framework including agent scheduling and task orchestration; Conduct post-training merging work such as data merging and OPD merging; Build and optimize the AI Agent memory system covering writing, organization, retrieval, and lifecycle management; Collaborate with product and engineering teams to deploy Agent Harness capabilities.

Sourced via tencent · Listed on CareerPlan, which tracks 845,000+ jobs from 20+ sources.