You own GlassDollar's company data as a product: what we collect, how good it is, what it costs and what we build next. You sit in the product team and report to our Product Owner.

Our data has grown faster than our ownership of it. Feeds run, but nobody owns whether they keep running. Most relationship records aren't classified yet, and some signals users ask for, like headcount changes, have no reliable live source. Fixing that is the job.

Who uses your work

People in vclOS	Search, lists, projects, PoC tracking, recommendations and weekly company updates.
Our AI agents & search	Clear definitions, provenance, confidence and freshness, measured on our search benchmark.
API and MCP partners	Stable definitions and fields to build on.
Marketing	Directory, rankings and reports through a quarterly data release.

When these conflict, the product wins.

What you own

• The datapoint backlog our engineers pull from, ranked by value, cost and achievable quality. You set priorities; engineers own building and running at scale.
• The data dictionary and schema as the source of truth.
• Signals: funding, acquisitions, shutdowns, case studies and headcount, with correct event dates and alerts when a feed stops.
• Data quality: coverage, accuracy, freshness and completeness, measured monthly.
• The data budget: vendors, APIs, scraping partners and scouting effort. You test sources on real samples before we commit, then negotiate.
• Human-in-the-loop work: what our scouting team collects and how it flows into the database.

You join sprint planning and review, triage data errors reported by users and agents, and prototype and ship datapoints yourself before handing over what needs to scale.

Your first 90 days
By day 90: relationship data is classified and measured, every feed has an owner and an alert, 3–5 new datapoints are live, and our search benchmark has moved against a target you set in month 1.

Month 1: baseline

• Data dictionary v1 for the top 25 datapoints, with source, owner, user and quality numbers.
• Every pipeline, vendor feed and scraping job inventoried with an owner, runbook and alert.
• Quick win: Scale partnership data point collection.
• Next year's data budget as a business case with a cost per datapoint.

Month 2: corporate–technology relationships

• Define corporations and relationship types: mention, partnership, PoC, implementation, investment, acquisition.
• Classify and clean existing records, report precision on a sample, and classify new records on arrival.
• Set a recall and lag baseline, then improve the relationship pipeline against it.
• Find a live acquisition source and keep acquired companies out of recommendations.

Month 3: new datapoints and agents

• Run 6–8 time-boxed datapoint experiments, each ending in ship, kill or park. Ship 3–5.
• Add provenance, confidence and freshness to the top 25 fields, and make customer data (lists, PoCs, notes) usable by agents with proper permissions.
• Decide where we go broad and where we go deep, and publish a kill list.
• Start a monthly data review with product and internal customers.

Who you are
Business-first, and technical enough to ship alone. You'd rather put a rough datapoint live this week than a perfect one next quarter.

Must-haves

• 5+ years with company, market or B2B data in a product, data or operations role, ideally at a data business, marketplace or SaaS company.
• You've owned an outcome with a budget and trade-offs inside a product team.
• Fluent Python: you write your own scrapers and enrichment scripts (requests or Playwright, pandas).
• Strong SQL on a production Postgres database.
• Hands-on with LLMs for extraction and classification, measured on a labelled sample.
• Practical entity resolution and deduplication.
• Working knowledge of GDPR for company and people data.
• You explain data trade-offs clearly to non-technical colleagues and leadership.
• Based in Berlin, or willing to relocate.

Nice to have

• German at reading level: DACH registers are a big part of our upside.
• Buying data from providers such as PitchBook, Crunchbase, Dealroom or company registers.
• Event or signal pipelines: change detection, event dating, alerting.
• Knowledge of the corporate innovation, VC or emerging technology ecosystem.

Stack: Postgres, Hasura, Python, PostHog and LLM APIs.

What this role is not:

• Not BI or analytics: no dashboards for their own sake.
• Not a data platform architect: engineers own infrastructure.
• Not ML research: you use models, you don't train them.
• Not customer workspace data (projects, departments, exports): that stays with product engineering.
• Not team-building from day one: you lead through scope and priorities.