About the role#
This role focuses on the bridge between research and practical applications. You will work on speculative decoding to improve accuracy and efficiency for our inference platform. You will own a project from data preparation through training and final evaluation. This involves working with core inference and Applied ML teams to integrate your work into our production environment.
What you'll do#
- Research and prototype new speculator algorithms.
- Experiment with architectural ideas and data curation to improve performance.
- Collaborate with customers to understand their specific needs.
- Learn how frontier inference systems operate in a production setting.
- Present your results to the team at the end of the internship.
What you'll need#
- Currently pursuing a degree in Computer Science, Electrical Engineering, Information Science, or a related field.
- Experience with Python and PyTorch.
- Familiarity with modern LLMs and generative models.
- Ability to perform effective hyperparameter searches and understand tuning trade-offs.
- Experience with SLURM or Kubernetes clusters and managing jobs in high-performance computing environments.
- Basic understanding of distributed training frameworks like FSDP or DeepSpeed.
- Meticulous attention to detail regarding data curation and model checkpoint evaluation.
- Publications at research conferences or journals such as NeurIPS, ICML, ICLR, ACL, EMNLP, or Interspeech.
Location & details#
- Location: San Francisco, California.
- Term: Winter 2027 (January 4th to April 9th).
- Modality: On-site.
- This is a full-time, paid internship.
- Sponsorship is not provided for this role.
About Together AI
Together AI operates a cloud platform for AI engineers and researchers. The company provides tools for inference, model shaping, and pre-training. Founded in 2022, it maintains its headquarters in San Francisco, California. The organization employs 415 people and serves clients ranging from AI startups to SaaS companies.
How to get in at Together AI
Securing an internship at Together AI requires speed, as early applicants are often reviewed before the candidate pool becomes unmanageable. Intern Insider sends an instant alert the moment a role matching your target is published anywhere, ensuring you apply among the first. This helps you avoid the frustration of submitting an application to a queue that has already grown too large to navigate. You can use this timing to your advantage to stay ahead of the competition. Beyond timing, reaching out to the right person often yields better results than a standard submission. Intern Insider surfaces the recruiters behind the roles at Together AI so you can contact them directly to ask about the position or a referral. This approach materially improves your response rates compared to waiting for an automated update. It is a more direct way to show your interest to the people actually managing the hiring process.
