Job Responsibilities：
1. Advanced post-training of large language models (e.g. SFT, RLHF/RLAIF, continual pretraining).
2. Aligning models for reliable JSON-schema function calls and external tool usage.
3. Design, deploy, and operate Model Context Protocol (MCP) servers that handle checkpoint routing, manage context windows, and enforce safety gates.
4. Experience in distributed training and inference with DeepSpeed/FSDP, LoRA/QLoRA, mixed precision, and performance tuning on vLLM or Triton clusters.
5. Build offline and live eval pipelines for alignment, factuality, grounding, and hallucinations.
Qualifications
1. Bachelor’s or Master’s degree in Computer Science, Artificial Intelligence, Machine Learning, or a related field.
2. 3+ years of experience in developing and optimizing large language models.
3. Proven track record in implementing advanced post-training techniques (SFT, RLHF, RLAIF, continual pretraining).
4. Hands-on experience with distributed training frameworks (DeepSpeed, FSDP) and optimization techniques (LoRA, QLoRA, mixed precision).
5. Familiarity with model alignment, JSON-schema function calls, and external tool integration.
6. Experience in building and maintaining evaluation pipelines for model performance assessment.
7. Proficiency in Python and relevant machine learning frameworks (e.g., PyTorch, TensorFlow).
8. Strong understanding of distributed systems and high-performance computing.
9. Experience with model deployment and inference optimization on vLLM or Triton clusters.
10. Knowledge of JSON-schema and API development.
