English

ZhiMian · AI Interview Assistant

Uses LLMs to analyse job descriptions and résumés, generating tailored questions, answer coaching, mock interviews and company research.

Get a free proposal and quote

Background

Job seekers face thousands of job descriptions, and generic question banks rarely match a specific role. ZhiMian combines JDs, résumés and LLMs into one loop — question generation, mock interviews, answer coaching and company research — while keeping deployment and troubleshooting manageable for a small team.

Solution

  • Structured question generation: parses the JD and résumé, then generates 5–10 tailored questions with assessment points, difficulty and suggested answer directions.
  • Live mock interviews: WebSocket streaming, multi-turn context, and an AI interviewer that follows up and gives immediate feedback.
  • Answer and résumé coaching: STAR-based rewrites in concise, standard and detailed versions, plus résumé match scores, skill gaps and ATS suggestions.
  • Company and role research: company background, interview process, common questions and salary ranges, with multi-role management and history.
  • Operable architecture: FastAPI with a PostgreSQL async stack, Docker Compose, Prometheus, Grafana, Loki, Jaeger, Alertmanager and Blackbox; self-hosted Ollama, JWT auth, plus content moderation and manual review.

Deliverables

  • User and admin front ends with a FastAPI backend
  • PostgreSQL data model, migrations and seed scripts
  • Six core prompt templates and the AI service layer
  • Docker Compose, Nginx, HTTPS and rollback scripts
  • Monitoring dashboards, tracing and operations documentation

Technology stack

FastAPIPostgreSQLDockerPrometheusGrafanaJaeger

Outcome

Feature-complete and in closed beta, pending filing approval before public registration. Production issues can be traced quickly through metrics, logs and distributed traces.

Services