We're running the pilot of a new embodied-AI data venture: a mobile app, a web back-office, and the cloud services connecting them, used by a small group of field collectors wearing data-capture devices. The core system is already built and largely working: real authentication, a task-claiming flow, resumable and verified cloud uploads, a review and settlement pipeline, an audit trail. You would take over technical ownership of that system, close out the remaining path to the first real collector payout, and direct one junior engineer currently on the project. You're expected to use AI coding tools (Claude Code or similar) heavily, but you're hired for judgment and ownership, not typing speed.

What you'd own Technical ownership of a live, mostly-built system: a React Native collector app, a back-office web console, and the cloud and ingest services behind them. You'll need to read and understand a real codebase quickly, not start from a blank page. The remaining critical path to launch, in rough dependency order: an Android native module for direct device communication (the biggest unknown, since the protocol has been captured but implementation hasn't started), local storage compatibility on the machines used to import data at collection sites, a security and permissions pass, getting the payment and payout flow live end to end, and running and fixing issues from a small internal pilot before any outside collector uses the system. The data-integrity invariants already built into the system, and holding the line on them going forward. Hardware self-reports data about its own recordings that cannot be trusted at face value, so duration and file structure get verified against raw evidence, never accepted from the device as-is. The same discipline applies to money: the app never sends a duration or an amount, only raw spans. The server alone computes settlement, and a retry can never produce a double payment. Direct technical mentorship of one junior engineer, delegating lower-risk, well-scoped work to them while keeping ownership of anything money- or integrity-critical yourself. Technical decisions in the absence of a dedicated PM. You'll work from a written specification but will need to make judgment calls, flag genuinely unresolved questions rather than guessing silently, and tell apart the blockers you can solve from the ones that depend on an external partner.

What we're looking for Senior-level experience shipping production systems that handled real user data or money, not just feature work inside an already-hardened codebase. Comfortable picking up someone else's working code and taking real ownership of it, including the parts you didn't write. Some exposure to native mobile development (Android), or willingness to go deep on a native module quickly. This is the largest remaining technical unknown on the project. A defensive-engineering instinct: default to distrusting an external input and verifying it, not trusting the fastest path to something that compiles. Comfortable directing and critically reviewing AI-generated code, not just using AI tools to produce it. You should be able to catch the specific failure mode where AI-written code looks correct and is subtly wrong. Full-stack range across mobile, backend and cloud services, and basic web front-end. Not required: machine learning or model-training experience. This is data and systems engineering, not ML.

Specific skills and tools React Native for the collector app, and React for the back-office console. This is the app's stack today. Native Android development, in Kotlin or Java. The device-communication module is a native module, and it is the single largest piece of unbuilt work. Backend and API development with a relational database and real transactional guarantees. The settlement logic depends on this: a retry must never produce a second payment. Experience with cloud object storage, ideally something S3-compatible. The current system uploads to VNG Cloud on GreenNode, with every part read back and checksummed after upload. Comfortable integrating third-party auth or payment providers. The app already uses Zalo for sign-in and ZaloPay for collector payouts. Direct experience with either is a plus, not a requirement. Git-based workflow with an existing CI setup. The project already runs nightly end-to-end tests, so you'd be working inside that, not building it from nothing. Heavy, daily use of AI coding tools such as Claude Code in a real production codebase, not just for personal projects or prototypes. Interview signal we're screening for: given a real, partially-unfamiliar codebase and a hardware component whose protocol has only just been captured, how would this person get from "I don't know this system yet" to shipping the remaining piece correctly, without breaking what already works.