MyInterview

MyInterview connects profile evidence, CV and job-description analysis, practice planning, text and voice interviews, scoring, feedback, recruiter workflows, and administration. The difficult work is not placing an AI chat box on a page; it is keeping those workflows consistent across product, data, runtime, and evaluation boundaries.

I lead the five-member graduation team and drive most of the product's technical execution across architecture, AI systems, backend, frontend integration, realtime interview flows, testing, and documentation. This describes primary contribution without presenting the project as solo work.

The system separates product rules, schemas, prompts, provider behavior, scoring, and persistence provenance. Candidate preparation remains text-first, while voice capabilities are introduced behind explicit readiness and provider gates. Structured output and business invariants are tested independently from prose quality.

This case study is grounded in private repository review, verification commands, runtime contracts, test surfaces, and project history. Public visuals reconstruct confirmed workflows with synthetic data; they do not expose real users, credentials, endpoints, or internal topology.

The project has reinforced that an AI product is only as useful as its failure handling, evidence trail, language policy, evaluation strategy, and cross-layer consistency. The strongest work is often the contract that prevents a confident model output from becoming an incorrect product decision.