Data Integration & Onboarding Intern (m/f/d)

About PRYZM Solutions

Pryzm Solutions builds AI-powered pricing intelligence for mid-market B2B manufacturers in the DACH region. We turn weeks-long manual pricing processes into data-driven, near-real-time decisions — EU-sovereign infrastructure.

About the Role

Every new customer we onboard comes with messy, heterogeneous data spread across ERP, CRM, and operational systems. Getting from "raw customer export" to "clean, model-ready data" is currently the single biggest bottleneck on our path to scale.

We're hiring our first Data Integration Intern to help us turn onboarding into a repeatable, documented, semi-automated process. Your work will directly shape how fast we can take on new customers — and how many we can serve at once.

What You'll Do

•        Standardize our data intake. Turn one-off onboarding workflows into a reproducible specification: what data we need, in what format, from which source systems.

•        Build a data quality assessment framework. Automate what our engineers do manually today — completeness checks, consistency validation, anomaly detection — into reusable Python notebooks and report templates.

•        Develop synthetic data generators. Build realistic pharma/manufacturing-style datasets we can use for demos, training, and testing without touching customer data.

•        Document onboarding runbooks. Shadow our onboarding work, extract the tacit knowledge, and turn it into step-by-step playbooks anyone on the team can follow.

•        Analyze pilot retrospectives. Where did we lose time? Which steps repeat across customers? Surface patterns that turn into product features.

What You Get

•        Paid monthly stipend

•        Certificate of completion

•        LinkedIn recommendation from the team

•        Real experience shipping at an early-stage AI startup

•        Direct mentorship from the technical team on causal AI modeling, and MLOps

•        A real shot at joining the core team full-time after the internship, based on performance

Location & Duration

•        Remote — EU time zones preferred

•        6 months, full-time preferred (part-time negotiable for strong candidates)

•        Occasional in-person meetups possible

Requirements

•        Strong Python skills, comfort with pandas and SQL

•        Obsessive attention to data quality — you notice when something is off

•        Clear written communication — you can turn a messy process into a clean document

•        Ability to work independently and ship without hand-holding

•        Genuine interest in B2B data, pricing, or industrial/pharma domains

Nice to have

•        Background in data science, computer science, industrial engineering, or a quantitative field

•        Exposure to ERP systems or CRM

•        Experience with ETL tooling (dbt, Airflow, Great Expectations) or synthetic data generation

•        German language skills

How to Apply

The application is the job itself. No CV needed.

We've prepared a messy, realistic dataset that mimics what a mid-market manufacturer would hand us on day one — inconsistent SKU IDs, missing costs, fragmented customer records. It's publicly available, no NDA required.

Your task:

1.      1. Download the dataset.

We use the UCI Online Retail II dataset as our test bed — a real, messy, two-year transactional dataset from a UK wholesaler (~540,000 rows). It has duplicates, cancellations, missing customer IDs, and inconsistent product codes: structurally similar to what we encounter when onboarding a new pharmaceutical or industrial manufacturing customer.

Download it here: https://archive.ics.uci.edu/dataset/502/online+retail+ii

2.      Your task is to treat this as if it were a Pryzm customer's first data handover. The domain is different — the data quality challenges are not. Run a data quality assessment. What's broken? What's missing? What would block a pricing model from training on this data?

3.      Propose an intake specification. If you were designing the form a new customer fills out to hand us this data, what fields would you require? What checks would you run automatically?

4.      Write it up. A 2–4 page document (PDF or Markdown) covering: your quality findings, your proposed specification, and one concrete improvement you'd prioritize first.

5.      Include at the top: your name, email, LinkedIn, and location.

Email your document to [email protected] with the subject line: Data Integration Intern — [Your Name]

If your work is sharp, we will schedule a call.