Senior Data Engineer: Entity Resolution & Reverse-Lookup Pipeline (Ai-native. Early-stage Startup)

We're an early-stage, remote-first people-search startup. Our product takes an identifier: a phone number, email, or name and resolves the person behind it, then expands into their connected information: associated names, locations and handles, other phone numbers and emails, relatives and associates, and related public records. We assemble this from licensed commercial identity data, public records, and phone/email-intelligence APIs into clear, readable reports. We already have a working MVP and detailed integration specs. We build AI-native: Claude Code, Cursor, and Codex are core to how we work, and we move fast because of it.

The core of the product is reverse-lookup and identity resolution: one identifier in, an accurate connected profile out — assembled primarily from licensed identity data, enriched with public records and phone/email signals. We're hiring a senior data engineer to build that engine. The make-or-break part is matching records from many sources into one accurate profile without ever merging two different people.

what you'll own

- a multi-source ingestion pipeline (vendor apis and licensed feeds), orchestrated as a cost-aware "waterfall": cheap filters first, charge-on-success enrichment next, premium sources gated last.
- The entity-resolution / record-linkage engine, and the logic that prevents false merges (the heart of the role).
- Name and address normalization, fuzzy matching, an identity and relationship graph, a daily refresh, and a search/serving layer.
- A suppression / opt-out stage (we honor removal requests).
- Per-source logging of cost, hit-rate, and latency so we can tune the waterfall.

must-have

1. Proven, hands-on entity resolution / record linkage (probabilistic, deterministic, or ML), and a real answer for how you prevent false merges. This is the non-negotiable core; please don't apply without it. Familiarity with Splink, Zingg, Senzing, dedupe, or equivalent.

2. AI-native development (required, not optional). You work daily in Claude Code, Cursor, or Codex and ship dramatically faster because of it. This isn't a box to tick: AI-accelerated delivery is how we operate, and it's central to how we'll evaluate you. Expect to walk us through your AI workflow.

3. Strong pipeline engineering on large, messy datasets; fuzzy matching; address normalization (e.g. Libpostal); Elasticsearch / Opensearch or similar.

strong plus — who we're really hoping to find

- field experience at a people-search, identity-verification, fraud-prevention, data-broker, or b2b data-enrichment company.
- Hands-on knowledge of the data-vendor landscape: US (e.g. LexisNexis / Accurint, Thomson Reuters CLEAR, Enformion / Tracers, TLOxp, IDI / idiCORE, Searchbug) and international (phone-intelligence, Pipl, sanctions, breach sources), plus the ability to advise us on which sources to prioritize and an honest, realistic view of where coverage is genuinely deep vs. Thin by region.

how we work

- project-based to start: a scoped production build, with strong potential to continue as we grow. Remote, flexible hours, senior rate.
- There's an existing AI-built prototype and detailed specs. Use them as a head start, or rebuild from scratch — your call. We care about the result, not preserving the code.

to apply

we ignore generic or templated proposals. Please answer these directly:

1. Describe an entity-resolution system you've built. What matching approach did you use, and specifically how did you prevent false merges?

2. With no unique ID linking records, briefly walk through your blocking and scoring logic.

3. No single data source covers every country. Given that, how would you architect regional coverage?

4. Which data vendors / feeds have you worked with, and which did you find reliable vs. Unreliable — and how did you judge?

5. Which AI dev tools do you use, and how do they change the way you build? Concretely, what's faster now that wasn't before?

Begin your proposal with the word "resolved" so we know you read this.