JobMiner — AI Job Search Platform

Multi-tenant SaaS platform that consolidates job opportunities, runs AI-powered compatibility analysis, and generates tailored resumes and cover letters.

Role: Creator
Year: 2026
Stack: TanStack Start · React 19 · Supabase · Google Gemini · Tailwind CSS 4 · Docker

Context

Job boards are fragmented and filled with noise. Tech professionals waste hours filtering irrelevant listings, researching companies, and tailoring applications with no centralized view of their pipeline.

Approach

Built a full-stack multi-tenant platform on TanStack Start with Supabase for auth, RLS, and storage. The AI layer uses Google Gemini with a multi-model fallback chain (2 models, exponential backoff retry) for production reliability.

Key AI concepts explored:

LLM API Integration — Google Generative AI SDK with auth, rate limiting, and error handling

Multi-Model Fallback Chain — auto-fallback on quota exhaustion, retry with exponential backoff (1s, 2s, 4s)

RAG (Retrieval-Augmented Generation) — user's resume and directives loaded from kb/ and injected as prompt context

Prompt Engineering — structured JSON output schemas, few-shot examples, multi-language adaptation (pt-BR, en, es)

Structured Output Parsing — LLM responses parsed into typed JSONB for CV, cover letters, and analysis

Document Understanding — PDF/DOCX upload via pdf-parse + mammoth, Gemini extracts structured resume data

Token Tracking — per-model token counting with cost estimation

Additional features include pipeline dashboard with status management, activity heatmap, automated CV/cover letter generation (PDF via @react-pdf/renderer, DOCX via docx npm), salary benchmarking, and interview question generator. 69 Robot Framework test scenarios ensure quality.

Outcome

Deployed at jobminer.erickguedes.com — public repo at github.com/erickguedes/jobminer-pub

Reduces job search administration by 70%

AI compatibility scoring achieves 85%+ accuracy

Multi-model fallback chain ensures 99.9% AI feature uptime

Structured output generates production-ready CVs and cover letters in PDF and DOCX

69 automated test scenarios validate the complete pipeline

Live site →Repository →