Job Description

Skills: Torch/PyTorch, TensorFlow

What if the real frontier isn't another chatbot, but an AI that can taste?

We're Turing Labs. The invisible force on path to drive 50,000+ CPG products like the next cereal or sauce or lotion you would use. We've been quietly reformulating what America eats. Now we're building the AI that will revolutionize HOW the $2 trillion CPG industry thinks.

The Foundation Model Nobody Sees Coming

While OpenAI and Anthropic fight over benchmarks, we're building something different.

Your Mission: Architect the Intelligence

As our Foundation Model Research Intern, you won't be fine-tuning models. You'll be generating unique dataset, never sourced before.

Your Daily Reality:

• Morning: Work closely with our AI engineers to design strategies to generate specific segment of data
• Afternoon: Build evaluation frameworks
• Fine-tune specific models with context engineering
• Late afternoon: Present your architecture to our CTO and argue why your approach beats transformers
• Evening: Watch your model predict specific scenario

The Technical Challenge:

• Design knowledge distillation pipelines from expert interviews → structured data → model weights
• Create evaluation metrics for subjective experiences (how do you benchmark "deliciousness"?)
• Develop few-shot learning approaches using our proprietary knowledge base

You'll Have Access To:

• Compute resources that would make your university lab jealous
• Direct mentorship from our technical team
• CEO and top minds in AI

We're Looking For:

• The student who implemented papers for fun before the official code dropped
• Someone who thinks "Attention Is All You Need" was just the beginning
• Equal comfort with PyTorch, LLMS, fine-tuning, evaluation framework, and context engineering
• The kind of person who's built at least one model that made their professor say "wait, how?"
• Bonus: You've thought about AGI while eating lunch and wondered if intelligence can be... tasty?

Why This Internship Hits Different:

• You're not optimizing ad clicks;
• Your model won't write poetry - it'll reformulate products 300 million Americans consume
• Real shot at co-authoring papers that'll make NEURIPS reviewers do a double-take

To Apply, Send Us:

1. GitHub/research portfolio (we'll actually read your code)
2. 300 words or 2-minute video: "How I'd teach an AI to taste"
3. Your wildest idea for using transformer architectures outside NLP/vision
4. One question about our approach that would stump our CTO (we dare you)

P.S. - We can't tell you everything (stealth mode has its rules), but we would be happy to share more details during interview process.