Product Engineer, Data (Full Stack, AI-Native)

Product Engineer, Data (Full Stack, AI-Native)

Location : Austin, TX preferred (hybrid); open to remote for the right candidate Type : Full time Reports to : VP of Engineering Compensation : $100,000 to $125,000 base salary, plus benefits and any applicable equity or bonus compensationAbout Living Security

Living Security is a B2B SaaS company in human risk management — we help large enterprises understand and reduce the human side of security risk. Our customers are major enterprise security teams, and our platform increasingly runs on AI: AI-generated content, risk scoring, and AI-native capabilities are core to where we're headed. We're a small engineering team with an unusually high output-per-engineer model, hiring builders who want that leverage.

About the Role

Our product is, at its core, a data product. Signals flow in from our customers' enterprise systems — identity providers, email, HR, security tools — get resolved to real people and teams, and feed the risk scoring that security leaders make decisions with. This role owns the front half of that machine: getting the data in, making it mean something, and keeping it correct.

We are hiring an engineer whose core work is data ingestion, integrations, entity resolution, and signal analysis. You'll build the connectors and sync jobs that pull data from customer systems, resolve messy real-world identity data into clean entities, analyze and shape the signals that drive scoring, and hunt down correctness issues across all of it. You'll work alongside the engineer who built much of this layer, take real ownership of major pieces, and become the second deep expert in the most important subsystem we have.

This is still a full stack role — you'll ship the APIs and admin/debugging surfaces around your pipelines, and you can work anywhere in our codebase — but the center of gravity is the data layer. Our data stack is Postgres, ClickHouse, and AWS, and we need you strongest exactly there.

How We Work

Every engineer here develops primarily through AI tools — Claude Code or similar. We don't write code by hand as our default; we direct, review, and validate. Your value is the judgment layer: knowing how data systems actually behave, where pipelines silently corrupt, what a suspicious distribution looks like — and using that to drive the AI hard and catch it when its output is plausible but wrong. In data work especially, plausible-but-wrong is the expensive kind of wrong.

We run continuous flow, not sprints: roughly one-week shippable deliverables, minimal ceremony between you and production. Success is measured in trustworthy data shipped to customers, not story points.

Be clear-eyed: this is a startup with hard enterprise commitments, and the pace reflects that. The team is small enough that there's no one to hand things off to. If you want a season of your career doing the most intense, highest-leverage work you've done, this is that. If you want a comfortable cruising altitude, it isn't.

What You'll Do

• Build and own data ingestion from customer enterprise systems: connectors, webhook handlers, sync jobs, historic backfills, and the rate limits, token lifecycles, and failure modes that come with pulling data from systems you don't control.
• Build integrations — both through unified API platforms (Nango or similar) and direct — and own them in production: when a customer's data stops flowing, you find out why and fix it.
• Own entity resolution: mapping messy, conflicting identity data from multiple sources into clean, correct entities — the people and teams everything downstream depends on. This is the hardest and most valuable data problem we have.
• Analyze and shape data signals: understand what incoming data actually means, derive the signals that feed risk scoring, and notice when a signal's behavior shifts in ways that would silently distort scores.
• Hunt and prevent correctness issues: write the SQL to find bad records, trace how they got that way, ship safe corrective migrations, and build the validation that kills the class of error.
• Build evaluation and regression coverage for the pipeline — the quality gates that let a fast-shipping team trust what's in production.
• Ship the full stack around your data work: the APIs that serve it and the internal admin and debugging surfaces that make it operable.
• Use Claude Code as your core workflow for design, implementation, testing, and validation — with your own data instincts as the quality gate.

What We're Looking For

• Real experience building data ingestion and integrations in production: pulling data from third-party systems via APIs and webhooks, handling auth lifecycles and rate limits, and debugging pipelines when they silently go wrong. Experience with unified API platforms (Nango or similar) is a plus.
• Deep Postgres, ClickHouse, and AWS — your strongest skills, non-negotiable. Postgres: schema design, query plans, indexing, safe migrations at scale. ClickHouse (or equivalent production columnar-store experience you can map directly): table engines, materialized views, knowing what belongs there versus the transactional database. AWS: the data-relevant services — S3, SQS/EventBridge, Lambda/ECS, RDS — and how production pipelines actually run on them.
• Excellent SQL. Not "comfortable with" — excellent. Complex joins, window functions, performance reasoning, and corrective work against production data are daily work here.
• Experience with entity resolution, identity data, or similar messy-real-world-data matching problems — or the demonstrated judgment to take on the hardest version of that problem fast.
• Full stack capability in our kind of stack (TypeScript, Node.js, Python, React): you can ship the APIs and internal tools around your pipelines without waiting on anyone.
• Instinct for data correctness: you notice when a distribution looks wrong, you distrust suspiciously clean results, and you validate before you trust.
• Daily, deep use of AI development tools (Claude Code strongly preferred). You don't write code by hand as your default — and you've developed the review discipline that makes that safe in data systems, where wrong answers look exactly like right ones.
• Experience building evals for AI/LLM outputs is a strong plus.
• Comfort with enterprise B2B realities: multi-tenant data isolation, large customer datasets, and data that customers' executives make decisions with.
• High ownership, low ego. You'd rather be the person who makes the data trustworthy than the person who's never blamed for it.

This is a mid-level role. We care about demonstrated pipeline-building and data judgment far more than years of experience or titles.

You'll Thrive Here If

• "The numbers look wrong" is a sentence that ruins your day until you've found out why.
• You like owning a system, not a slice — and watching it get more correct, faster, and more trusted over time.
• You want to work where data engineering touches the product directly, not three teams removed from users.
• You want a stretch of your career defined by maximum output and learning, around people operating the same way.
• You use AI tools systematically and aggressively, with your own judgment as the quality gate.

This Role Is Not For Someone Who

• Wants a pure analytics or notebook role — this is production data engineering.
• Is uncomfortable owning integrations with systems they don't control — third-party APIs breaking is normal Tuesday here.
• Trusts AI output without validating it — in data work, that's how silent corruption ships.
• Needs a fully specified data model before starting.
• Wants big-company pace and predictability. We're a startup in a sprint, and we're honest about it.

Interview Process

Three steps, designed to evaluate real working ability without days off or speculative travel:

1. Screen (45 min, remote). How you think about data systems: pipelines you've built on Postgres/ClickHouse/AWS, a data corruption story you root-caused, and how you use AI tools daily in data work. Expect specific questions about query plans, columnar-store tradeoffs, and pipeline architecture on AWS.
2. Live working session (2 hrs, remote). Investigate a realistic data problem in our stack using Claude Code — a pipeline producing subtly wrong numbers. We watch how you drive the AI, how you validate against the data itself, and whether you catch the plausible-but-wrong diagnosis.
3. In-person final (half day, Austin). Meet the team and make sure it's a fit, both ways. Travel covered for remote candidates.

For candidates between roles or contracting, we can structure the final step as a paid 30-day contract-to-hire — an option, not a requirement.

Location

Austin, TX hybrid is our strong preference; Austin-area candidates should work in person on a regular cadence. For the right candidate we're open to US remote, with occasional travel to Austin.

How to Apply

Tell us about a pipeline you built — ideally on Postgres, ClickHouse, and AWS — and a time the data was wrong in a way nobody noticed at first — how you found it, fixed it, and made sure it stayed fixed. We especially want to hear how you use Claude Code or similar AI tools in data work, including a time the AI's answer was plausible but wrong and how you caught it.