ML Training & Evaluation Intern

Upsilon Holdings; Hours; On-site; EUR 6,700 - 11,000; Full time

Job summary

Upsilon Holdings is offering an internship focused on ML evaluation, data prep, prompt/retrieval experiments, and dashboards to understand model behavior in real usage. You’ll run evaluation harnesses, support experiments, and translate ML results into product decisions, while improving pipeline reliability with logging and documentation.

The role welcomes curiosity and detail-oriented learners. Required: Python basics, data handling, curiosity about model evaluation, and strong attention to

Qualifications

• Experience with Python and data analysis.
• Interest in ML evaluation and model behavior.
• Strong attention to repeatable results.

Responsibilities

• Run evaluation harnesses and track quality changes over time.
• Support prompt, retrieval, and dataset iteration experiments.
• Help produce concise reports translating ML results into product decisions.
• Improve reliability of pipelines: logging, consistency checks, and documentation.

Skills

Python basics

Data handling

Model evaluation curiosity

Attention to detail

Python debugging

Tools

Vector search

Embeddings

Dashboards

Notebook workflows

LLM tooling basics

Job description

Role overview

You’ll work on data preparation, prompt/retrieval experiments, benchmark runs, and dashboards that help us understand model behavior in real usage. This is practical ML: measurable, iterative, and tied to product outcomes.

Responsibilities

• Run evaluation harnesses and track quality changes over time
• Support prompt, retrieval, and dataset iteration experiments
• Help produce concise reports that translate ML results into product decisions
• Improve reliability of pipelines: logging, consistency checks, and documentation
  
  

Requirements

• Comfort with Python fundamentals and working with data
• Curiosity about model behavior and evaluation methodology
• Strong attention to detail (repeatability matters here)
  
  

Nice to have

• Experience with vector search / embeddings
• Dashboards or notebook workflows
• Basic LLM tooling familiarity
  
  

Internship compensation

Internships can be paid or unpaid depending on scope, location, and legal constraints. In the UK, unpaid arrangements are typically only feasible in limited situations (for example: genuine volunteering with charities/non-profits, short work-shadowing/observation, or certain structured course-related placements). If the role involves productive work that benefits the company and looks like a “worker” role, it generally must be paid at least the National Minimum Wage. We will always structure internships compliantly.

Note: This is general information and not legal advice. Final structure depends on the role and jurisdiction.