AI Resume Tailoring · SaaS2026
Paari.online
Full-stack job application suite featuring AI LaTeX resume tailoring, isolated compilation sandbox, automatic job matching, and PDF parsing.
- Role
- Full Stack Engineer — AI LaTeX tailoring engine, sandboxed compilation worker
- Timeline
- 2026 · Active
- Stack
- NestJSTypeScriptReactPrismaPostgreSQLRedisBullMQGemini AILaTeX / TeX LiveAWS S3RazorpayTailwind CSS
01
Screens & Architecture
02
The Problem
Job seekers waste hours manually tailoring resumes to Job Descriptions (JDs) and struggle with ATS rejections, while existing AI tools destroy custom LaTeX formatting and preamble macros during optimization.
03
How I built it
- AI LaTeX Engine: Integrated Google Gemini AI to selectively extract and tailor safe text sections (skills, summary, experience) while strictly preserving preamble code, custom commands, and layout structure.
- Sandboxed LaTeX Compilation Queue: Designed an async compilation pipeline using Redis and BullMQ to run isolated TeX Live builds without blocking NestJS main API threads, uploading outputs to AWS S3 via presigned URLs.
- PDF-to-LaTeX Fallback: Built a PDF parsing pipeline with pdf-parse that converts raw unformatted resume text into Paari's standardized, editable LaTeX template.
- Monorepo Architecture & Tiered SaaS: Built a Turborepo monorepo sharing type packages between web & API, complete with Razorpay billing, coupon management, and plan-gated version history.
04
Outcome
- Reduced resume tailoring time from hours to <15 seconds with 100% LaTeX macro and preamble preservation.
- Achieved non-blocking, isolated document generation and PDF rendering with zero server thread stalls using asynchronous BullMQ workers.