InterLyt AI Interview Platform

Problem Being Solved

In resume screening and mock interview workflows, the hard part is not just calling an LLM, but connecting resume parsing, interview Q&A, and knowledge-base retrieval into a runnable, debuggable application. InterLyt targets this kind of full-stack AI interview system, covering resume parsing for PDF and DOCX, interview question generation from resume content, and RAG-based knowledge-base Q&A. For engineers, it is useful for studying how Spring AI, PostgreSQL, Redis Streams, and React fit together.

How the Skill Works

Resume management: after upload, resumes move through an asynchronous analysis pipeline from PENDING to PROCESSING, then COMPLETED or FAILED; the flow supports retries, hash-based duplicate detection, and PDF report export.
Mock interviews: questions are generated from the resume, with configurable follow-ups and a batch evaluation strategy that reduces long-context token pressure before producing a summary report.
Knowledge-base Q&A: uploaded documents are chunked and vectorized, then served through pgvector retrieval and SSE streaming responses.
The system relies on PostgreSQL + pgvector, Redis Streams, MinIO or S3-compatible storage, and Spring AI 2.0; troubleshooting usually starts with configuration, Redis connectivity, vector table initialization, and PDF Chinese font handling.

Scope and Limitations

InterLyt is closer to a self-hosted full-stack project scaffold than a single API. It expects a database, cache, object storage, and an LLM API key. If the goal is only one-shot interview question generation, a direct model call may be simpler; if the goal is deployable resume analysis and interview records, this skill gives module boundaries, async task flows, and deployment troubleshooting paths.