# 顶尖应届-语音大模型算法研究员-MiMo

External opening: you cannot apply through AI Native Match; apply on the official site.

Apply through the official channel. The full external receipt stays in your workspace. When signed in, three minimal event types explicitly recorded through the CLI are reported by default: selection, self-reported application and channel observation. Browsing or saving a role does not report an event; external applications do not appear under My applications.

- Company: 小米 (Stage unknown)
- Locations: 北京
- Role: Algorithm
- Apply on the official site: https://xiaomi.jobs.f.mioffice.cn/toptalent/position/7646711916187289892/detail

## Role description from source

1、大规模语音模态预训练
研究如何用千万小时级别语音数据在百亿级别参数模型上进行高效预训练，有效提取与利用通用声学与语义特征，提升语言理解及语音生成的一致性和自然度；
2、多语言语音理解与生成
研究跨语言语音数据的共享表示方法，提升语音模型对多语言、方言的适配能力；
3、噪声环境及复杂声学场景下的处理能力
研究语音大模型在嘈杂、混响、远场等典型场景下的泛化性；
4、探索高效语音信息压缩方法
研究语音模态压缩方法，以实现长时理解，并适配不同类型的设备（云侧和端侧）的部署需求。
任职要求

接口顶部原始条件（未做资格推断）
{"city_list": [{"code": "CT_11", "en_name": "Beijing", "i18n_name": "北京", "location_type": 3, "mdm_code": "MDCY00000001", "name": "北京", "node_status": 1, "py_name": "beijing"}], "job_post_info": {"HighlightList": [], "JobChannelPublishList": [], "address_list": [], "city_list": [], "correlation_job_list": [], "job_level_id_list": [], "job_post_object_value_map": {}, "job_post_process_time_list": [], "recruitment_type": {"active_status": 1, "children": [], "depth": 2, "en_name": "Regular", "i18n_name": "正式", "id": "201", "name": "正式", "parent": {"active_status": 1, "children": [], "depth": 1, "en_name": "Campus", "i18n_name": "校招", "id": "2", "name": "校招", "selectability": 1}, "selectability": 1}, "storefront_list": [], "tag_list": [], "target_major_list": []}, "process_type": 2, "project_type": 2, "recruit_type": {"active_status": 1, "children": [], "depth": 2, "en_name": "Regular", "i18n_name": "正式", "id": "201", "name": "正式", "parent": {"active_status": 1, "children": [], "depth": 1, "en_name": "Campus", "i18n_name": "校招", "id": "2", "name": "校招", "selectability": 1}, "selectability": 1}}

---
opening_id: ff5d0ff3-744a-4bc3-adcb-7cc568db1327

`anm openings save --id ff5d0ff3-744a-4bc3-adcb-7cc568db1327`