HR AI / AI Interviewer

Context

A private, work-in-progress recruiting platform built with a Next.js dashboard and FastAPI backend. Companies can create jobs, share tokenized application links, collect resume uploads, run chat-based AI screening interviews, and review structured candidate evaluations. The system includes tenant-aware data access, cookie-based sessions, CSRF protection, rate limiting, account hardening, audit logging, email notifications, async database migrations, and pluggable language-model providers.

Role and scope

Independent full-stack project: designed and built the multi-tenant recruiting workflow, AI interview service, employer dashboard, public candidate flow, and security foundations.

Challenge

Early-stage hiring teams need a consistent way to collect candidate context and run an initial screen without stitching together job forms, resume folders, interview notes, and evaluation spreadsheets. The project explores how to turn that workflow into a reviewable product while keeping hiring decisions with people.

Approach: Built a multi-tenant employer dashboard for jobs, candidates, interviews, evaluations, and team access; added public tokenized application and interview flows; and implemented resume validation, persisted transcripts, structured evaluation data, email invitations, and recruiter review surfaces. Security boundaries include HttpOnly session cookies, CSRF checks, rate limits, password controls, audit events, and file validation.

Key decisions and trade-offs

• Used tokenized public links for candidate flows so applicants can participate without receiving employer dashboard access.
• Kept tenant ownership explicit in the data and service layers so jobs, candidates, interviews, and users stay scoped to the correct company.
• Made LLM access provider-configurable so interview behavior can be evaluated across providers without coupling the product to one vendor.
• Stored transcripts and structured evaluations separately so recruiters can review the evidence behind an AI-assisted assessment.

Outcome

• Implemented an end-to-end workflow from job creation and candidate application through AI screening and structured review.
• Added security and abuse-resistance controls around authentication, uploads, public links, rate limits, and tenant isolation.
• Created a foundation for recruiter review while keeping AI output as decision support rather than an automatic hiring decision.
• Added backend and frontend tests around authentication, interview flows, candidate evaluation, and security-sensitive behavior.

Stack

Stack used in this system · Independent full-stack project