- Company Name
- ByteDance
- Job Title
- Student Researcher (Doubao (Seed) - LLM Foundation Research)- 2025 Summer/Fall/Winter (PhD)
- Job Description
-
**Job Title:** Student Researcher – LLM Foundation Research (PhD) 2025 Summer/Fall/Winter
**Role Summary:**
Conduct advanced research and development on large language models (LLMs), focusing on post‑training techniques, reinforcement learning from human feedback, interpretability, reasoning, data synthesis, tool use, and multi‑agent systems. Work flexibly (onsite or remote, part‑ or full‑time) to design algorithms, build scalable oversight, and evaluate model performance for applications such as search, recommendation, advertising, content creation, and customer service.
**Expectations:**
- Deliver novel RLHF algorithms, reward‑modeling methods, and oversight mechanisms that improve LLM robustness, generalization, and alignment.
- Enhance model interpretability and reasoning capabilities across the entire development pipeline.
- Produce high‑quality multimodal data and advanced decoding strategies for complex task solving.
- Publish research findings in top‑tier conferences and collaborate effectively with interdisciplinary teams.
**Key Responsibilities:**
- Design and implement reinforcement learning algorithms (e.g., heuristic‑guided search, multi‑agent RL) for LLMs.
- Formulate and test new reward‑modeling approaches to boost accuracy and generalization.
- Develop scalable oversight frameworks for monitoring large‑scale LLMs.
- Improve interpretability of model decisions and outputs.
- Advance reasoning and planning techniques, including system‑2 thinking, MCTS, A* decoding, and tool‑use agents.
- Synthesize and augment large‑scale multimodal datasets for pre‑training, SFT, and RLHF stages.
- Create robust evaluation methodologies to diagnose model capabilities and drive iterative improvements.
- Build and test single‑ and multi‑agent systems that interact with APIs, code interpreters, and external tools.
- Publish research results and present findings to internal and external audiences.
**Required Skills:**
- Strong theoretical and practical knowledge of LLMs, reinforcement learning, NLP, and machine learning.
- Proven research record with publications at venues such as NeurIPS, ICML, ACL, EMNLP, etc.
- Proficient Python programming; solid data structures and algorithm fundamentals.
- Experience with competition coding (e.g., ACM/ICPC, USACO, TopCoder, Kaggle) preferred.
- Excellent communication and teamwork abilities; ability to explore and integrate emerging technologies.
**Required Education & Certifications:**
- Currently enrolled in a PhD program in Computer Science, Linguistics, Statistics, or a related technical field.
- No additional certifications required; relevant research, internship, or competition experience is a plus.