Research Intern, Frontier Speculative Decoding (Summer 2027)
Core
Research and prototype new efficient speculative decoding algorithms, draft model architectures, and training recipes to enable faster generation on frontier LLMs without changing outputs.
Role type
Research Intern (Frontier Speculative Decoding)
Builds
Draft models and inference algorithms for production serving engines
Domain
AI/ML, Large Language Models, Inference Systems
Deliverable
production ML models
Required skills
Machine Learning, Deep Learning, Distributed Training, Model Fine-tuning, Knowledge Distillation, Experiment Design, LLM Architectures, Serving Systems
Preferred skills
Speculative Decoding, Reinforcement Learning, Open Source Contributions, Long-context Handling, Agentic Workloads
Technologies
Megatron-LM, FSDP, vLLM, SGLang, TensorRT-LLM, DeepSeek, Kimi, GLM
Responsibilities
Research and prototype new efficient speculative decoding algorithms and draft model architectures; Train, fine-tune and distill draft models, and design rigorous experiments to validate hypotheses; Implement optimized versions for serving engines to show efficacy in real world workloads; Communicate project plans, progress, and results to the broader team; Share findings through technical write-ups and blog posts, and help bring successful ideas into production
Seniority
Intern
