Must Haves
- 3+ years of hands-on data engineering / backend engineering experience, including designing and operating production data pipelines (extraction, transformation, scheduling, monitoring).
- Strong SQL and relational database experience, ideally with Postgres specifically — comfortable writing and reviewing complex views, and reasoning carefully about correctness (e.g., avoiding fan-out double-counting, picking the right amount/date field) rather than trusting a query at face value.
- Practical experience building with LLMs / AI applications — e.g., building tools or agents on top of Claude, OpenAI, or similar; familiarity with the Model Context Protocol (MCP) or an equivalent tool-calling framework is a strong plus, or willingness to learn it quickly.
- Salesforce data experience — comfortable extracting data via API from Salesforce (objects, fields, relationships), even if you're not a Salesforce admin/developer by trade.
- Cloud infrastructure competency — able to independently stand up and manage a small cloud environment (compute, managed database, scheduled jobs) on a major provider (AWS/Azure/GCP).
- Comfortable working independently with light daily oversight — this is a small, senior-heavy team; you should be able to take a scoped workstream and run with it, surfacing questions rather than needing close direction.
- Strong written communication — this engagement lives or dies on documentation and runbooks; you should be able to write clearly for a non-technical audience as well as a technical one.
- Comfortable with ambiguity and evolving scope — this is discovery-and-build work for a client who is still defining their own requirements; definitions and priorities will be captured and refined during the engagement, not fully specified upfront.
Nice to have
- Prior experience working in or with the nonprofit / philanthropy sector.
- Experience building "curated tool" or semantic-layer approaches to grounding LLMs in structured data, as opposed to open-ended natural-language-to-SQL.
- Experience with geographic/public data analysis (e.g., joining external datasets by ZIP/district/geography).
- Prior consulting or client-facing delivery experience, even if this role itself is not client facing day-to-day.
Day-to-day
- Stand up the data platform. Provision a cloud environment and managed Postgres database; build and schedule ingestion pipelines from multiple sources (a Salesforce CRM extract, a public-data source, and a large corpus of unstructured documents from Google Drive), each landing in a consistent, documented schema.
- Build a curated, verifiable query layer. Translate business metric definitions (provided by the client and by the engagement's architecture lead) into reviewed SQL views and a small set of parameterized, enum-constrained tool signatures — each validated against a known-correct answer. This is deliberately not a general text-to-SQL system; precision and auditability matter more than flexibility.
- Build and deploy a working MCP (Model Context Protocol) tool server. Wire the curated query layer to Claude via a small set of production-quality tools, and demo them live against real client data — a genuine working slice of a larger 2027 answer-layer build, not a mockup.
- Build a public-data proof-of-concept. Analyze and join an external public dataset against the client's CRM data to demonstrate what outside context adds to program evaluation, with honest, clearly stated limits on what the analysis can and can't support.
- Connect and demo Claude against the client's own CRM data, walking through a tiered set of real questions to show what a simple LLM connection can answer today and precisely where it hits its limits.
- Profile and audit the unstructured document corpus — volume, format, and condition — producing the data that lets the next project phase be scoped and priced with confidence.
- Document everything for handoff. Write plain-English runbooks and a data dictionary so the client's own team is never dependent on any single person to maintain or extend what you build.
- Work daily with, and demo weekly to, Precocity's AI & Architecture lead, who sets architecture direction and reviews your work — especially anything that will be shown live to the client — before it ships.