AI Researcher (Intern)
IRIS8 trains specialist models on customers' own data and runs them inside their environments.
Join as an intern to fine-tune and evaluate open-weight models, run benchmarks, help build data pipelines, and support private deployments.
About Iris 8
Organisations are renting AI from frontier labs. Their data leaves the building, costs keep climbing, and they own nothing they pay to improve.
Iris 8 changes that. We build sovereign AI: private, specialist models trained on an organisation's own data, running inside its own perimeter, with every request routed to the smallest model that gets the job done.
Founders have built multi million Dollar companies in the past.
About the role
You will work with the founding team on the research behind our private models: fine-tuning and evaluating open-weight LLMs, building evaluation and data pipelines, exploring distillation, and helping take models from experiment to a private deployment. You will own well-scoped pieces of that work, and you will learn how we turn research into something a customer can run in their own environment.
What you'll do
- Fine-tune and evaluate open-weight language models on customer or internal datasets
- Run and improve benchmarks that measure accuracy, cost and safety
- Help build and maintain data pipelines for training and evaluation
- Support distillation of larger models into smaller specialist models
- Help deploy and test models in private environments: on-prem, private cloud or air-gapped
- Write up results clearly so the team can decide what to ship next
You may be a good fit if you
- Are pursuing or recently completed a degree in computer science, machine learning, or a related field
- Have hands-on experience with Python and at least one ML framework, such as PyTorch
- Have coursework, research, or project experience with LLMs, fine-tuning, or model evaluation
- Are comfortable reading papers and turning an idea into a reproducible experiment
- Can write clearly about what you tried, what happened, and what you would do next
- Want to work on models that run in a customer's own environment, not only in a notebook
Nice to have
- Experience with open-weight models, LoRA, or similar parameter-efficient fine-tuning
- Familiarity with evaluation harnesses, distillation, or inference serving (vLLM or similar)
- Experience building data pipelines for training or evaluation
- Interest in private or on-prem deployment, or work in a regulated setting
Why join
- Ground floor at a company built by founders who have done it before
- Work directly with the founding team on models that ship into customer environments
- Hard problems: frontier-quality AI at a fraction of the cost, inside the world's most regulated environments
How to apply
Send your resume, plus a link to something you've built, to [email protected].