Research Intern, Frontier Speculative Decoding (Summer 2027)
Together AISan Francisco, USposted 6h ago
- PayUSD / yr
- $121k – 146k
- Where
- San Francisco, US
- Posted
- Oct 16
- 6 hours ago
- Type
- Internship
- Setup
- On-site
At a glance
Research novel speculative decoding algorithms, train and distill draft models on GPU clusters, and validate them in real serving workloads to make frontier LLM generation faster and more efficient.
Summarized by AI from the original posting
What you'll do
- Research and prototype efficient speculative decoding algorithms and draft model architectures
- Train, fine-tune, and distill draft models
- Design rigorous experiments to validate hypotheses
- Implement optimized serving-engine versions
- Evaluate ideas on realistic serving workloads
- Communicate project plans, progress, and results
- Share findings through technical write-ups and blog posts
- Help bring successful ideas into production
Requirements
- 01Strong knowledge of machine learning and deep learning fundamentals
- 02Hands-on experience training or fine-tuning large models with distributed training frameworks
- 03Familiarity with modern model architectures
- 04Prior research experience in speculative decoding, efficient inference, or efficient machine learning preferred
- 05Publications at leading machine learning or systems conferences preferred
- 06Open-source machine learning contributions preferred
Perks
Skills
Full description
About the Role
The Speculative Decoding team at Together AI builds the draft models and algorithms that let large language models generate faster without changing their outputs. Our speculators run in production on frontier open models such as DeepSeek, Kimi and GLM. Our work sits at the intersection of model training, inference algorithms and serving systems, and we co-design all three to lower the cost and latency of generation.
As a research intern, you will dive into open questions in speculative decoding: new drafter architectures and verification schemes, knowledge distillation and training recipes that align draft and target models, and drafters that adapt as traffic shifts. You will train and fine-tune draft models at scale on dedicated GPU clusters, and test your ideas against realistic serving workloads, including long-context and agentic traffic.
Projects aim to answer important research questions in speculative decoding, enabling faster generation on frontier models, drafters that keep pace with changing workloads, and robust, reproducible evaluation under real traffic.
Past Research
Speculative decoding research by Together researchers, interns and academic collaborators includes:
- Speculative Speculative Decoding (ICLR 2026)
- When RL Meets Adaptive Speculative Training: A Unified Training-Serving System (ICML 2026)
- Beat the Long Tail: Distribution-Aware Speculative Decoding for RL Training (MLSys 2026)
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 version for serving engines to show efficacy in real world workloads
- Communicate the plans, progress, and results of projects to the broader team
- Share findings through technical write-ups and blog posts, and help bring successful ideas into production
Requirements
- Currently pursuing a Master's or Ph.D. degree in Computer Science, Electrical Engineering, or a related field
- Strong knowledge of Machine Learning and Deep Learning fundamentals
- Hands-on experience training or fine-tuning large models with distributed training frameworks (e.g., Megatron-LM, FSDP)
- Familiarity with modern model architectures, such as Mixture-of-Experts and hybrid attention
Preferred Qualifications
- Prior research experience in speculative decoding, efficient inference, or efficient machine learning
- Publications at leading conferences in machine learning or systems (e.g., NeurIPS, ICML, ICLR, MLSys)
- Experience with knowledge distillation, post-training, or reinforcement learning
- Familiarity with LLM serving engines (e.g., vLLM, SGLang, TensorRT-LLM)
- Contributions to open-source machine learning projects
About Together AI
Together AI, the AI Native Cloud, is purpose-built for AI engineers. AI application developers get high-performance inference that scales reliably, fine-tuning and reinforcement learning for creating frontier-level specialized models, and pre-training at massive scale for fully custom intelligence, all around a marketplace of leading open models that teams can run, adapt, and own. Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom, Together serves 400+ trillion tokens a month.
Internship Program Details
Our internship program runs 12 to 14 weeks, giving you the opportunity to work alongside industry-leading engineers and researchers across multiple teams. This cohort's internship dates span May 17th to August 6th or June 14th to September 3rd.
Compensation
We offer competitive compensation, housing stipends, and other competitive benefits. The estimated US hourly rate for this role is $58 - $70 an hour. Our hourly rates are determined by location, level and role. Individual compensation will be determined by experience, skills, and job-related knowledge.
Equal Opportunity
Together AI is an Equal Opportunity Employer and is proud to offer equal employment opportunity to everyone regardless of race, color, ancestry, religion, sex, national origin, sexual orientation, age, citizenship, marital status, disability, gender identity, veteran status, and more.
For more information about our privacy policy, please visit: https://www.together.ai/privacy