AI Testing Assistant

Mosaic Conseil operates as a small, remote workshop dedicated to rebuilding one business artifact at a time. Whether it is a three year plan, a pricing grid, or a diligence file, we take things apart on the bench and put them back together against the evidence of the invoice ledger. The process is personal and focused. We do not work on standing retainers and we do not produce slide decks. Instead, we return the same artifact you sent, redrawn by hand and supported by a written argument. Our work is handled on paper and backed by a telephone number that answers for ninety days. We keep the workshop small on purpose to ensure that the people who do the work also answer their own phones. There are no meetings we did not schedule and no corporate theater. We focus on four specific rooms: financial services, industrial networks, growing software companies, and healthcare systems.

About Mosaic Conseil
We are a boutique consultancy that values intellectual honesty and the discipline of a small, focused team. By staying small and fully remote, we remain independent and focused on the evidence at the floor of the business. Our clients come to us when they need a business plan or a financial model rebuilt with extreme precision. We provide a workspace where the primary goal is getting the numbers and the logic right, rather than making things look good in a presentation. Every file that leaves our workshop carries two names on the cover: one person who builds and one who argues against what was built. This ensures that every assumption is tested before a file closes. This environment is built for thinkers who prefer a written argument over a status meeting.

What you would do
As an AI Testing Assistant, your work focuses on the intersection of business logic and automated analysis. You will be responsible for testing the outputs of AI models used to process financial and industrial data. This is an entry level role where you support our consultants by stress testing the artifacts we build. Your day to day work involves:

• Reviewing AI generated summaries and data extractions for accuracy against original source documents like ledgers or contracts.
• Running standardized tests on internal AI tools to identify where they fail to grasp complex financial logic or industrial constraints.
• Building simple test cases to verify that a pricing model or a business plan remains consistent when variables are changed.
• Documenting specific instances of model error or hallucination in a clear, written format.
• Assisting the challenger in our builder and challenger model by providing evidence of where an automated calculation may have deviated from the truth.
• Comparing different AI model outputs to see which one handles specific industry terminology from healthcare or industrial networks more reliably.
• Verifying that the data used for a specific business artifact has been correctly mapped by our automated scripts.

Helpful background
We are looking for individuals who have a natural curiosity about how technology works and, more importantly, how it breaks. This role is accessible to those starting their career or looking for part time work while finishing their studies. Helpful traits and experiences include:

• Basic familiarity with Large Language Models and an understanding of how to prompt them for technical tasks.
• Competence in using spreadsheets like Excel or Google Sheets to verify calculations.
• Strong writing skills and the ability to explain a technical error in plain English.
• A background in business, finance, or computer science is useful but we value the ability to follow a logical argument above a specific degree.
• Technical comfort with remote work tools and a disciplined approach to managing your own schedule.
• Experience with personal projects or academic work that involves data analysis or AI testing.
• An interest in the specific sectors we serve, particularly financial services and industrial networks.

How we work
Our workshop is entirely remote, meaning you can work from wherever you are most productive. We do not have an office and we do not hold meetings unless they are strictly necessary for the work. We communicate primarily through written documents and internal notes. Our builder and challenger model means that your work as a tester is central to our reputation. You are not just a support person: you are an essential part of the verification process that allows us to stand behind every artifact we produce. We prioritize deep work and give our team the quiet they need to find the errors that others miss. There is no requirement for residency in a specific city or country, as we operate a global, decentralized bench.

What the role offers