Enterprise analytics teams are drowning in an ad-hoc question queue: "what was churn last quarter, but by the finance definition, not the sales one?" Copilot guesses. Snowflake's assistant only sees Snowflake. Neither of them will stop and ask the person who actually owns the definition.
ThinkPal takes that queue off the analytics team across the whole estate — Snowflake and Power BI together — and when a question can't be answered from what's already agreed, it escalates to a named business owner instead of making something up. Every answer carries a record of who decided the definition, when, and on what evidence. We serve those answers into whatever assistant the org already uses (Claude, ChatGPT, Copilot) over MCP.
We're a seed-stage company with a technical founder in the Bay Area, design partners in financial services and healthcare, and no engineering team yet. You would be the first hire. You'll own large pieces of the product outright, and the architecture decisions you make in the first year will still be load-bearing at 25 engineers.
You'll spend most of your time on the core system: the semantic layer that reconciles conflicting definitions across Snowflake semantic views, Power BI semantic models, and dbt projects; the agent loop that turns a natural-language question into a governed query or, when the definition is missing or disputed, into a routed request to the right human; the attestation ledger that records every decision with owner, timestamp, and evidence; and the MCP server that exposes all of it to Claude, ChatGPT, and Copilot.
Concretely, in the first six months, that means shipping importers for Power BI semantic models and Snowflake semantic views, building the conflict-detection and escalation workflow that design partners are already asking for, hardening the customer-owned read path so it runs inside a regulated buyer's perimeter, and standing up the eval harness that tells us when a model upgrade changes answer quality. You'll sit in design-partner calls, hear the objection directly, and ship the fix.
You've shipped production software inside a real enterprise data estate — at a bank, an insurer, a large SaaS company, or a data platform vendor — and you have opinions about semantic layers because you've been burned by three of them disagreeing. You're fluent in at least one of Snowflake, Power BI/DAX, or dbt, and comfortable enough in the others to be dangerous. You've built with LLMs beyond a demo: you know what an eval suite is for, why temperature zero doesn't mean deterministic, and how to make an agent refuse rather than guess.
You write clean, boring, well-tested code in Python and TypeScript (or can get there fast), and you're at home with SQL that runs against billions of rows. You'd rather ship something small to a real customer this week than something perfect next quarter. You work well with a founder who is three or four time zones away, which means writing things down, making decisions in the open, and pushing back when something's wrong.
Nice to have, not required: experience with MCP or other tool-calling protocols, prior work in a regulated industry (finserv, healthcare), experience being an early engineer at a startup, and a track record of open-source contributions in the data tooling space.
Python, TypeScript, Postgres, Snowflake, Power BI (TMDL/DAX), dbt, MCP, Claude and OpenAI APIs, AWS. You'll have a say in most of what comes next.
We're building one engineering hub, and we're choosing between Toronto and São Paulo based in part on where the first hire lands. You should be based in one of those cities or willing to relocate to one; we'll hire you locally through a compliant employer- or contractor-of-record and move you onto a local entity as we grow. The role is remote-first with regular overlap with Pacific time.
Compensation is competitive for senior engineers in your market, and as employee #1 the equity is a meaningful founding-level grant, not a standard early-employee option. We'll share specific numbers in the first conversation. Standard benefits per local norms; hardware and home-office budget; travel to the Bay Area a few times a year.
Four steps, two weeks end to end: a 30-minute conversation with the founder, a paid take-home built on a slice of the real problem (a couple of conflicting semantic models and a question that can't be answered from either), a 90-minute technical deep-dive on your solution and your past work, and a final conversation about the company, the equity, and whether we'd enjoy building this together. We'll tell you where you stand after every step.
Send a short note about the worst semantic-layer disagreement you've ever had to resolve, plus a link to something you've built, to contact@getthinkpal.com. No cover letter needed.
Senior QA Engineer (Canada)
Zazz
ETL Tester with Salesforce Healthcare
Cloud Data Vision
Enterprise CLoud Architect
Net2Source (N2S)
Senior Test Analyst
VStream Labs Inc.
Test Analyst - Senior
Maracas Bay Consulting Inc
Associate Technical Architect - ML
Quantiphi