Data Science Consultant

The Data Science Consultant works on several client projects in parallel, from raw data through to the final presentation. Day to day, this means:

Collecting, cleaning, and preparing data of varying quality, mainly with Python and Power Query.
Building reports, dashboards, and automations in Power BI and the Microsoft Power Platform.
Developing predictive and machine learning models based on what each client needs.
Designing and building AI agent pipelines that automate processes in production.
Owning the full deliverable, including the visualizations and the presentation of results.
Explaining scope, limitations, and results clearly to both technical and business audiences.
Contributing to proposals, requirements gathering, and scoping of new projects.
Documenting the work so that anyone on the team can reproduce it.

Success in the role looks like clients coming back for a second project, and deliverables that keep running after we hand them over.

Qualifications and requirements

A degree in Computer Science, Engineering, Economics, Mathematics, or another quantitative field.
2+ years of experience in data science, analytics, or a related area.
Demonstrable experience with Power BI, Power Query, and the Power Platform, with at least two projects where you built the solution end to end: source connections, data transformation, and reports in production.
Experience designing AI agent pipelines that reached production.
Working knowledge of Python for data science, including pandas, NumPy, and scikit-learn. This is assessed during the technical test.
SQL, plus familiarity with cloud data platforms such as Databricks, Snowflake, BigQuery, Azure, AWS, or GCP.
Experience working with messy or low quality data, including preprocessing and statistical analysis.
Advanced professional English. This one is a hard requirement, since most client work is in English.
The ability to organize a workload, deliver on time with minimal supervision, and give early warning when timelines are at risk.
Honesty about what a model can and cannot do. We would rather hear the limitation early than read an overstated result.

See also