Назад

Не получаете ответ?

Telegram-вакансии старше 7 дней могут быть уже неактуальны.

8 дней назад

Algorithm Engineer (LLM Agent)

Формат работы
onsite
Тип работы
fulltime
Грейд
trainee
Английский
b2
Страна
China
vacancy_detail.hirify_telegram_tooltipВакансия из Telegram канала -

Мэтч & Сопровод

Покажет вашу совместимость и напишет письмо

Описание вакансии

TL;DR
Algorithm Engineer (LLM Agent) (AI/LLM): Developing core LLM agent capabilities, evaluation workflows, and post-training experiments with an accent on planning, tool calling, context and memory management. Focus on harness design, execution trajectory analysis, RAG and prompt strategies, fine-tuning, SFT, and DPO.

Algorithm Engineer (LLM Agent) - Campus Recruitment

Company

01.AI

Conditions

2 days ago

Skills

Candidate Availability

Required and preferred rules are kept separate and reflect the wording in the original posting.

About the Role

You will work on core LLM Agent algorithms including planning, tool calling, context and memory management. You will explore modern Harness design, build and maintain evaluation data and experiments, analyze execution trajectories and failures, explore RAG, prompt and context strategies, fine-tuning and post-training such as SFT and DPO, and integrate validated solutions with product and engineering teams.

Requirements

  • Class of 2027 graduate with master's degree or above in AI, computer science, statistics, mathematics or related field
  • Solid machine learning, deep learning and algorithm foundations; understands Transformer and LLMs; proficient in Python and PyTorch or TensorFlow
  • Understands Agent mechanisms and modern Harness design
  • Experiment design and problem analysis ability
  • Proactively applies Agent and Harness technology; Agent, RAG, Harness or fine-tuning projects or research preferred
  • Good communication and English technical documentation reading; open source, writing or competition experience preferred

Responsibilities

  • Develop core Agent capabilities including task decomposition, planning, tool calling, context and memory management
  • Explore and improve Harness design and evaluate impact on success rate, reliability, efficiency and cost
  • Build and maintain Agent evaluation data and experiment workflows and locate bottlenecks
  • Analyze execution trajectories and failure cases and design targeted experiments
  • Explore RAG, prompt and context strategies, fine-tuning and preference alignment, and join SFT and DPO experiments
  • Collaborate with product, algorithm and engineering teams to integrate validated solutions and iterate on feedback

Будьте осторожны

Если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого. Это мошенники. Обязательно жмите «Пожаловаться» или пишите в поддержку. Подробнее в гайде →

Текст вакансии взят без изменений

Источник -