3 months ago
Sŏngnam, Korea, SouthMid Level
Responsibilities
- Analyze model performance bottlenecks, failure cases, and training objectives with research and model-training teams to define data requirements.
- Design, generate, refine, and curate instruction, preference, reasoning, domain-specific, and synthetic training data.
- Build and operate automated data-generation pipelines and large-scale data-production workflows.
- Evaluate data quality and training impact using quantitative and qualitative criteria, LLM-as-a-Judge, rule-based validation, and human feedback.
- Develop iterative generation and filtering strategies and optimize pipelines for cost, throughput, and quality.
Requirements
- At least 3 years of experience in LLM, machine learning, or data generation work.
- Broad understanding of deep learning, machine learning, and natural language processing.
- Understanding of training-data composition, preprocessing, quality evaluation, and incorporation into model training.
- Proficiency in Python-based data processing and automation development.
- Experience processing, refining, filtering, and managing the quality of large-scale training datasets, or equivalent capability.
- Experience generating synthetic data with LLMs, evaluating data, or curating data.
- Understanding of or experience with LLM APIs and prompting or generation strategies.
- Strong problem-solving, collaboration, and communication skills.
- Preferred: experience building large-scale LLM training data for pre-training, supervised fine-tuning, DPO, RLHF, or RLAIF.
- Preferred: experience with OpenAI Evals, LM Evaluation Harness, DeepEval, or a proprietary evaluation system.
- Preferred: experience designing challenging LLM data involving multi-turn dialogue, tool calling, agents, reasoning, or code generation.
- Preferred: experience building Python-based data pipelines or large-scale data-processing environments with Airflow, Ray, or Spark.
- Preferred: contributions to open-source Generative AI, LLM, or NLP projects, or publications and technical presentations.
Benefits
- The hiring process includes document screening, a coding or assignment test, a first video interview of approximately one hour, a second in-person or video interview of approximately three hours, and offer discussions.
- The process and schedule may change depending on circumstances, with results communicated individually by email.
- Veterans and individuals eligible for employment protection are given preference under applicable laws.
- Applicants with registered disabilities are given preference under applicable law.
- A three-month probationary period may apply.
- A reference check may be conducted with the applicant's consent after the interview process.
Tech Stack
Categories
About 42dot
42dot builds software-defined vehicle and autonomous mobility platforms for automakers and mobility service providers. Its products include AI-driven vehicle software, data-based update services, and core platform components such as API gateways and identity/access management to connect services securely. The privately held company is headquartered in Seongnam-si, Gyeonggi-do, and focuses on transportation and automotive markets.