Backend and Data Systems Engineer

Vepler is an AI-native property intelligence company. We combine research and technology to help organisations understand the built environment and make informed decisions.The opportunityVepler's platform is built from public records about the UK built environment: planning, environmental risk, land, energy, addressing and more. Those records are fragmented, contradictory and constantly changing. The systems that bring them in at national scale, resolve them to the right entities and keep the result current are what the company runs on. This role owns a core part of them.You would join a small engineering team, working directly with the founders. You make the architecture calls in your domain, deploy your own work, and decide what the platform does next there. As the team grows, you shape how it works.We use AI throughout: an inference fleet we run ourselves, agentic pipelines that keep national coverage growing, and an evaluation discipline that measures everything before it ships. You would build that infrastructure.What you will doOwn the ingestion systems that keep the platform current across its national sources: reliability, throughput, recovery and the tooling around themOwn the entity resolution that ties each incoming record to the right address, site or property, and raise coverage across the estateBuild geospatial datasets from source acquisition through to a versioned product customers query, with PostGIS and the pipeline tooling around itMake the architecture decisions in your domain and see them through to productionSet the standard for how systems are tested, measured and operated, and document it for the engineers who followWhat we are looking forNode.js and TypeScript in production, with real systems behind youSQL well past the ORM: query plans, indexes, and what a write costs at scaleData pipelines built for change: retries, idempotency, deduplication, backfills and change detectionPractical geospatial work with PostGIS or similar: spatial queries, coordinate systems, geometry validationDiagnosing production problems on your own and fixing them properlyClear written English, particularly about failure and uncertainty. Most of how we work is writtenFluency with current AI tooling, and the judgement to check its output. Python is useful. We will teach you the rest of our stackWhat you getA rate set against the senior market, reviewed at a fixed point, and a completion payment on accepted delivery and handoverRemote-first, flexible hours, and an agreed overlap with the UK teamDirect access to the founders and authority over the decisions in your domainThe tools we use ourselves: current frontier models, our own inference fleet, and the budget to run what the work needsWhere this goesThe role starts as an initial engagement with a review at a fixed point. If both sides want to continue, it becomes a long-term position through local employment: private health insurance from day one, 25 days of leave plus public holidays, a learning budget, equipment and coworking, and a stake in the company, with terms set out before you accept.By six months you would own an end-to-end domain and take part in hiring. By twelve to twenty-four months you would be a core technical owner of the platform, on either a senior technical track or a technical leadership track as the team grows.How we workRemote-first, with an agreed overlap window rather than fixed hoursEvidence over opinion: everything ships measuredWritten by default: decisions, trade-offs and post-mortems live in documentsSmall and direct: engineering talks to the decision-makerHow we hireFour stages inside two weeks, a decision within 48 hours of the last one, and no unpaid work.A short written exchange by email: three questions on real problems from our estateIntroductory conversation (45 minutes): your background, ambitions and the workWorking session, paid at a day rate: a real problem on sanitised data, in your own time, with whatever tooling you normally useWalkthrough (45 minutes): we change the requirement and work on your solution together #J-18808-Ljbffr