Hired: A Distributed Clean-Architecture Platform for Automated Career Intelligence and Conversational Interview Coaching

Information asymmetry between software engineering job seekers and hiring entities creates acute market inefficiencies, particularly in emerging tech sectors. Existing job aggregation platforms provide search capabilities without actionable feedback, while commercial mock-interview coaching remains inaccessible to recent graduates. This working paper details the architectural design, implementation, and empirical evaluation of Hired, an end-to-end career intelligence and interview coaching system. The platform couples a React 19 single-page application with an asynchronous FastAPI microservice tier. The data plane combines a concurrent scraping engine aggregating postings across five heterogeneous job sources (RemoteOK, We Work Remotely, Arbeitnow, Greenhouse API, JSearch) with an idempotent normalization pipeline. The intelligence tier implements ten domain services powered by Groq-accelerated LLaMA 3.3 70B inference, enforcing strict JSON schemas, domain persona injection, and defensive output parsing. Both tiers are structured according to Clean Architecture, decoupling core domain logic from UI and infrastructure drivers. We present system latency benchmarks, validate the resilience of structured prompt extraction against LLM drift, and discuss engineering trade-offs regarding SQLite file locking under concurrent writes and client-side viseme lip-sync synchronisation for 3D conversational avatars. Preprint, not peer reviewed.