About This Role: We have already decided to put AI in front of our data. This role decides whether we can trust what it says.
REP's analytics direction is MCP-first: our internal agents, ChatDash (internally built, conversational BI platform), and Claude now query our Microsoft Fabric data platform and other sources directly. Agents are only as accurate as the definitions underneath them. In addition to building pipelines to model data from sources into Fabric, the AI-Native Analytics Engineer builds the semantic models and context layers that let both people and AI agents get accurate, grounded answers from REP's enterprise data.
Where a traditional analytics engineer stops at clean models for dashboards, this role extends the work into the agentic control plane: the certified metric definitions, data skills, evaluation suites, and guardrails that AI depends on to query data and act without hallucinating. And you will do it the AI-native way - using AI coding assistants to take new data from bronze to gold, with generated tests and automated reconciliation, compressing work that traditionally took a data engineer, a QA tester, and rounds of business UAT into one seat. What AI cannot do is the judgment: the spec, the review, and the call on whether the answer is correct. That is the job.
Just as important: at REP, the business builds first. Our end users create working versions of the analytics they need on their own - in Claude or directly in ChatDash - often before certified data or clear semantics exist for that domain. You will not be writing requirements documents and building from scratch. The requirements are the thing the business has already built and is already using. Where certified data and semantics exist, your job is to make sure they are readily accessible through Claude, ChatDash, and other development approaches, so people build on trusted ground from the start. Where they do not, your job is to take the working version and figure out how to get its data flowing systemically, with clear semantics and guardrails, without breaking what the business made.
This is a senior individual contributor role with real ownership. You will own REP's metric definitions of record, develop data pipelines according to a medallion architecture, direct our external data engineering partner as the technical owner, and work alongside our AI Solutions Engineer - they build the agents and tools; you build the governed data those agents rely on.
Essential Functions: Include the following. Other duties may be assigned.
Business-Built Solutions (Harden) - Treat working versions built by business users in Claude and ChatDash as the requirements: reverse-engineer the questions, logic, sources, and implied metric definitions they encode
- Take those working versions and get their data flowing systemically - replacing ad-hoc extracts, uploads, and one-off queries with governed Fabric pipelines and certified semantics
- Preserve what the business built while hardening it: the user's working version keeps working, now on trusted data with clear definitions and guardrails
- Make certified data and semantics accessible wherever people build - Claude, ChatDash, REP MCP, and other development approaches - so new work starts on governed ground
- Partner with the business builders directly, confirming that the hardened version answers the same questions they were answering before, and explaining any differences
Pipelines and Data Products (Build) - Build Fabric pipelines and gold data products end to end with AI-assisted development: ingestion into bronze, cleansing and conforming in silver, and gold tables and views ready for reporting and agents
- Generate test suites and automated reconciliation against source-system reports, replacing most of the manual UAT burden on the business
- Derive the spec from the working version: the business builder is the named consumer, their use is the decision, and their current outputs become the acceptance criteria - with freshness targets and a business owner confirmed before productionizing
- Keep transformation code version-controlled, tested, and documented well enough that another engineer - or an agent - can extend it without asking
Semantic Layer and Context Engineering (Trust) - Deliver a certified semantic layer over REP's Fabric gold layer - metrics, entity definitions, join paths, and grain rules - that REP MCP and ChatDash query by default, starting with inventory and net sales
- Establish one definition of record per metric, retiring or explicitly mapping conflicting legacy definitions across Power BI, Polar, and NetSuite
- Structure business rules, metadata, and documentation as durable working memory for agents, not after-the-fact commentary
- Resolve data-trust escalations at the root - in definitions and lineage - rather than through one-off reconciliations
AI Answer Accuracy (Evaluate) - Stand up an evaluation framework: a question bank with known-correct answers and automated regression checks run on every semantic-layer or model change
- Publish a regular accuracy scorecard so leaders and the AI Governance committee know where agent answers can be trusted - and where they cannot
- Catch regressions before release, not after a business user reports them
Governance Guardrails (Protect) - Implement access controls and row- and column-level policy in the semantic and MCP layers so agents only return data the requesting user is authorized to see
- Define which agent actions are read-only versus write-capable, and review write paths against REP's connector governance framework before exposure
- Maintain auditable lineage from any agent-delivered answer back to source, sufficient for Finance and governance review
Vendor Direction (Own) - Act as technical owner of our external data engineering partner: write build specs precise enough to execute and measure against
- Review delivered work against acceptance criteria and reject what misses the bar - professionally and with evidence
- Keep vendor scopes bounded to vertical, testable work so REP is never dependent on a vendor to understand its own data
What You Will Bring: To perform this job successfully, an individual must be able to perform each essential duty satisfactorily. The requirements listed below are representative of the knowledge, skill, and/or ability required. Reasonable accommodations may be made to enable individuals with disabilities to perform the essential functions.
- A hardener, not a gatekeeper: You see a business user's working prototype as a gift, not a mess. You can read what they built, understand the intent, and harden it without taking ownership away from them.
- A semantic and context engineer: You know when a problem belongs in the semantic layer versus a prompt, application logic, or an ad-hoc query - and you can defend the line. You can tell a definitional problem from a grain problem from a source-system problem.
- Verification-minded: You assume AI-generated output is wrong until tested, and you build the test before trusting the result. You say so plainly when confidence is low.
- AI-native, not AI-dependent: You treat AI coding assistants as a velocity multiplier under review, not an authority. You read and test what they produce before it ships.
- A governance owner: You apply governance to your own work without being asked, make the conservative call on data exposure when scope is ambiguous, and hold the architecture gates under pressure.
- A technical owner of partners: You translate business needs into specs a vendor can execute, and you are comfortable telling a partner their work does not meet the bar.
Education and/or Experience: - Bachelor's degree (B. A.) from four-year college or university
- 5+ years in analytics engineering, data engineering, or a closely related role, operating as a senior individual contributor
- Written high-performance SQL and Python dataflows against a cloud warehouse - Microsoft Fabric preferred; Snowflake, Databricks, or BigQuery experience transfers
- Built and maintained models in a medallion (bronze/silver/gold) architecture with dbt-style transformation, testing, and documentation discipline
- Designed or owned a semantic layer or metrics layer that served as the single source of truth for business metrics
- Used AI coding assistants (e.g., Claude Code, Cursor, Copilot) to materially accelerate data pipeline delivery
- Built or contributed to evaluation frameworks for LLM or text-to-SQL accuracy, or exposed governed data to AI agents via MCP or similar interfaces (a strong plus)
- Worked across messy source systems such as NetSuite, Shopify, Amazon SP-API, or HubSpot without having the model handed to you
- Implemented row- and column-level security, access policy, and lineage in a production data platform
- Taken analyst- or business-built prototypes (spreadsheets, notebooks, BI workbooks, or AI-generated reports) and turned them into production data products
- Directed external data engineering partners through specs and acceptance testing
Core Competencies: Action Oriented - Taking on new opportunities and tough challenges with a sense of urgency, high energy, and enthusiasm.
Resourcefulness - Securing and deploying resources effectively and efficiently.
Collaborates - Building partnerships and working collaboratively with others to meet shared objectives.
Self-Development - Actively seeking new ways to grow and be challenged using both formal and informal development channels.
Semantic & Context Engineering - Designing metric definitions, join paths, and context that let agents query data deterministically and accurately.
Technical Craftsmanship - Writing tested, documented, production-grade SQL and Python across the modern data stack.
Verification & Intellectual Honesty - Treating accuracy as an ongoing, measured system and communicating its limits before they surprise anyone.
Governance & Risk Ownership - Making agent data access safe by default and holding the line without supervision.
Productionizing Business-Built Work - Turning working prototypes into governed, systemic data products while keeping the builder in the loop.
Partner Direction - Specifying, reviewing, and accepting or rejecting vendor work as REP's technical owner.
What's in it for You: Our commitment to quality and transparency applies not only to our products, but to our people. We are focused on creating a fun, exciting, collaborative space that's centered around health and well-being. We empower everyone on our team to take control of their careers and balance their work and life in a modern, fast-paced environment.
We offer an
exceptional compensation package which includes:
Compensation base range: $120,0000 - $140,000 annually with bonus potential
Day One Benefits! - Medical, Dental, Vision (Competitive Benefits Packages Available)
- Health Savings Account (HSA) with employer contributions
- Flexible Spending Account (FSA) options
- LTD/STD, Life and AD&D (100% premiums covered by REP)
- 401k and Roth options with employer match up to 4%
- Employee Assistance Program (EAP)
Recharge Yourself! - 4 weeks of PTO accrued during your first year
- 48 hour
