Web/app engineer + search engineer

Search Engineer About the role We're looking for a search engineer to own the card identification and matching layer of a graded-card pricing tool. When someone scans a card's certificate number, the system has to work out exactly which card it is — the right set, the right variant, the right grade — by resolving against several external data sources. Everything downstream depends on that answer being correct. Getting it right is the difference between a trustworthy valuation and a confidently wrong one. This is a long-term role. The identification layer is central to the product, so the work will range across accuracy improvements, new data source integrations, search relevance and ranking, evaluation tooling, and performance. You'll be the person who owns matching quality as the product grows. What you'll start on The first project is a focused accuracy push: measuring how often the system prices the wrong card or fails to price a real one, then closing the gaps. It begins with building an evaluation harness against a dataset of around 2,000 real scans and recording a baseline, so that later improvements can be demonstrated rather than asserted. That work runs roughly two weeks and will give you a thorough grounding in how the matching cascade behaves. What we're looking for Solid TypeScript and Node.js. The codebase is a Next.js application, but you'll spend most of your time in the library and API layers rather than the interface. Real experience with fuzzy matching or entity resolution — scoring candidates across multiple attributes, setting acceptance thresholds, and knowing when the right answer is to reject a match rather than guess at one. This is the heart of the job. An instinct for search relevance. You should recognise, for instance, why results ranked by popularity instead of relevance will systematically misidentify obscure items. Comfort with measurement: precision and recall, confidence calibration, and designing metrics that hold up when someone questions them. A lot of this role is knowing whether a change actually helped. Test-driven development. The codebase is strict about it — tests are co-located, written first, and never hit the network. Willingness to dig into data. Often the fastest way to prioritise is to check how often a suspected problem actually occurs in production, rather than reasoning about it. Nice to have Experience with evaluation harnesses or regression suites built on golden datasets. Deterministic test fixtures, particularly recording and replaying third-party API responses. Working against external APIs with rate limits and retry behaviour. Familiarity with Vitest. Card or collectibles knowledge isn't needed. The matching problems are general ones. How we work Feature branches and pull requests, conventional commits, no new dependencies without discussion. We'd rather you ask questions early than build on an assumption. When something can't be measured with the data available, we want to hear that upfront along with what would fix it — not discover it at the end. If you can't do this, Pls don't apply. - Can you work on US timezone? If so, from when to when? - Can you reply within 10~15mins? - When finished the day, should report to me what you have done. - This is Milestone work. I will pay ~$400 per week. long-term project. ~6months - Should start from today - You should complete the task day by day ------------------- If you are ok, pls share your resume.