LLM Pretraining Data Technical Staff (Agentic Models)

About the Role We are seeking a LLM Pretraining Data Technical Staff to design and build high-quality datasets for large language model training for agentic and reasoning-heavy tasks. The role focuses on data curation, synthetic data generation, and task trajectory construction. Responsibilities

1. 01

   Develop and generate synthetic tasks and datasets for LLM training, including:

2. 02

   Verifiable tasks (e.g., coding task solving, mathematical problem solving)

3. 03

   Non-verifiable tasks (e.g., open-ended reasoning and general problem solving)

4. 04

   Construct and scale data synthesis workflows for domains including agentic tasks, coding tasks, math problems and general reasoning.

5. 05

   Contribute to the development of agentic task environments and evaluation setups for LLM training.

Basic Qualifications 1.2+ years of experience in large language model developement.

2. 02

   Familiarity with LLM training workflows, including pretraining, mid-training and post-training.

3. 03

   Hands-on experience with agentic model training and evaluation.

4. 04

   Experience designing evaluation tasks for agentic, reasoning, coding, or math benchmarks.

5. 05

   Strong programming skills (commonly Python).

6. 01

   Familarity with reinforcement learning algorithms and implementation for LLM training.

7. 02

   Hands-on experience building sandbox environments for task verification (e.g., code execution environments).