LLM Engineer (Reinforcement Learning)

We are looking for the best 🏆

LLM Engineer (Reinforcement Learning) 🤖

LLM Engineer(Reinforcement Learning) designs LLM learning pipelines and trains generative language models that can be utilized in real services. You will contribute to the continuous improvement of quality by constantly trying new methodologies, launching services that are essential to real users, and refining LLM so that it can improve its own quality.

Responsibilities 📝

• Improving the efficiency of the LLM learning process: Overall efficiency improvement of the learning process using Direct Alignment Algorithm / PPO, GRPO, DPO, etc. of PLM or Fine-tuned LLM.
• Improving the overall accuracy and stability of the generated results: Designing a learning structure that prevents Reward Hacking and enables Self-Refine to improve the quality of the generated results.
• Developing a basic model that can be linked with external knowledge and APIs: LLM learning that selects the necessary external linkage tool on its own depending on the type of instruction.

Qualifications 🎓

• 5+ years of experience in Deep Learning or NLP.
• Proficient programming (Python & pytorch) skills.
• Experience in model design, training, evaluation, and optimization using PyTorch.
• Ability to learn and troubleshoot LLM using GPU.
• Experience in distributed learning frameworks (Slurm, DDP, Horovod, etc.).
• Ability to collaborate smoothly with colleagues.

Preferred Qualifications 🌟

• Deep Learning/NLP related paper submission or Master's or Doctoral degree holder.
• Paper presentation experience at major academic conferences (ACL, EMNLP, NeurIPS, etc.).
• Experience in Docker and Kubernetes: Experience in designing and managing learning pipelines using GPU clusters.
• Experience in learning and service development using GPU: Experience in building GPU-based Training or Inference systems.
• LLM's Post-training related experience: Experience in utilizing Supervised Fine-Tuning and Parameter Efficient Fine-Tuning.

Interview Process 💼

• Document screening - Coding test - Video interview (within 1 hour) - Face-to-face or video interview (within 3 hours) - Final pass.
• The selection process may vary depending on the job, and may change depending on the schedule and situation.
• The interview schedule and results will be sent individually to the email you registered in the application form.

Additional Information ℹ️

• When submitting your resume, please exclude information prohibited by the Recruitment Procedure Act, such as resident registration number, family relations, marital status, annual salary, photos, physical condition, and place of origin.
• Please upload all submitted files in PDF format less than 30MB. (If you have any problems uploading your resume, please send your resume and the URL of the position you wish to apply for to [email protected].)
• A reputation check may be conducted with the applicant's consent after the interview process is complete.
• National veterans and employment protection beneficiaries are given preferential treatment in accordance with relevant laws and regulations.
• Persons with disabilities are given preferential treatment in accordance with the Act on the Promotion of Employment of Persons with Disabilities and Vocational Rehabilitation.
• 42dot does not accept unsolicited resumes from search firms and does not pay fees for unsolicited resumes.

※ Please check the following information before applying.

• The way 42dot works, 42dot Way →
• 42dot's unique work immersion program, Employee Engagement Program →