Resumate AI — Conversational AI Resume Builder & ATS Optimization Platform
Conversational AI resume builder, document ingestion, and ATS optimization platform with real-time A4 preview and server-side PDF export.
My Role
Full-Stack Developer & AI Systems Engineer
Category
Full-Stack AI SaaS
Timeline / Date
2026
Primary Stack
Next.js 16 & React 18
01 / The Problem
The Operational & Engineering Challenge
Traditional form-based resume builders overwhelm job seekers by requiring manual formatting, complex section organization, manual ATS keyword adjustment, and regional convention knowledge. Non-technical candidates struggle to translate their career history into modern, recruiter-ready formats, while static editors fail to provide instant feedback or context-aware rewriting.
02 / The Solution
Architecture & System Implementation
Engineered Resumate AI, a conversational full-stack SaaS platform built with Next.js 16, Google Gemini AI, Supabase PostgreSQL, and Playwright. Users can converse in Bangla, Banglish, or English or upload existing resumes (PDF, DOCX, TXT, scanned images), and the system extracts, rewrites, and renders the data across 8 regional templates in real time, backed by server-side headless Chromium A4 PDF export.
03 / Platform Showcase
Conversational Resume Platform Gateway & Onboarding
Modern full-stack web application landing interface welcoming job seekers with options to upload existing CV documents (PDF, DOCX, TXT, images) or start interactive conversational creation with trilingual AI assistance.
04 / Key Features
Core System Capabilities & Workflows
Conversational multi-turn CV creation and editing in Bangla, Banglish and English.
Feature 01
Resume ingestion from PDF, DOCX, TXT and scanned image files using document parsing and Gemini Vision.
Feature 02
8 professional regional CV templates for Bangladesh, Global ATS, International, MNC, Germany/DACH, Nordic Europe, Australia/New Zealand and Europass-style applications.
Feature 03
Real-time A4 CV preview with natural-language styling, section reordering, photo control and one-page optimization.
Feature 04
Atomic Undo/Redo, version history, multiple CV management and quiet database auto-save.
Feature 05
Secure server-side Playwright/Chromium PDF generation with Supabase PostgreSQL persistence and protected download authorization.
Feature 06
05 / Conversational AI & Design Suite
Conversational AI Workspaces & Template Suite
Purpose-built interfaces demonstrating multimodal document ingestion, real-time conversational editing in Banglish, global ATS layout conversion, and 1-page fit A4 document export.
Multimodal Ingestion & Modern Layout
Initial AI conversion gateway formatting candidate career history, contact details, and core skills into a structured two-column layout.
Conversational Experience Tuning in Banglish
Interactive split-screen interface where users converse in Banglish or Bengali to enhance bullet points with action-driven metrics and impact.
Global ATS Layout Optimization
Deterministic conversion into an ATS-compliant single-column layout engineered for applicant tracking system parseability and typography clarity.
1-Page Fit Active & Polished A4 Output
Algorithmic vertical whitespace balancing maintaining strict 1-page fit alongside server-side Playwright PDF download authorization.
08 / Engineering Contribution
My Specific Role & Responsibilities
Full-Stack Developer and AI Systems Engineer responsible for the complete application architecture, multi-model Google Gemini integration, real-time React state synchronization, Supabase PostgreSQL schema with Row Level Security (RLS), Playwright server-side PDF rendering, and production Docker deployment on Render.
09 / Technology Architecture
Full-Stack Technology Decisions
Backend & APIs
- Next.js 16 API Routes
- Node.js
- Playwright Chromium
- JWT Auth
Frontend & UI
- React 18
- TypeScript
- Tailwind CSS
- Lucide Icons
Database Layer
- Supabase
- PostgreSQL
- Row Level Security (RLS)
- Schema Migrations
Integrations & Tools
- Google Gemini API
- Gemini Multimodal Vision
- OpenAI API (Fallback)
- bKash Gateway Architecture
- Docker / Render
10 / Data Layer
Database Architecture & Relational Schema
Engineered on Supabase PostgreSQL with strict Row Level Security (RLS) policies, providing secure multi-tenant resume isolation, atomic version snapshots, and guest-to-user account migration.
Key Relational Entities
- users: Registered user credentials, profile metadata, and password hashes
- resumes: Master CV documents storing JSONB resume data, design configs, and payment states
- resume_versions: Historical state snapshots for rollback, undo/redo, and change tracking
- conversations & chat_messages: Multi-turn AI interaction logs and diff previews per CV
- payments: Server-verified transaction records strictly tied to verified resume IDs
Schema Design Principles
- PostgreSQL Row Level Security (RLS) isolating user records and HMAC-signed guest sessions
- JSONB document columns enabling agile schema evolution without breaking data integrity
- Automated foreign key cascading ensuring clean lifecycle management on CV deletion
11 / System Flow
API Architecture & Data Flow Lifecycle
A robust 8-stage data pipeline transforming raw inputs into validated JSON resume data, dynamic React views, and pixel-perfect server-side PDF streams.
12 / Problem Solving
Engineering Challenges & Technical Solutions
Challenge 01
Preventing LLM hallucination and schema drift during multi-turn conversational edits.
Engineered Solution:
Enforced structured JSON MIME outputs with Google Gemini, backed by an automated 2-stage retry validator and custom runtime normalizers that discard invalid fields.
Challenge 02
Processing mixed-script and colloquial Bengali/Banglish commands for resume modifications.
Engineered Solution:
Designed a dual-layer prompt architecture paired with regex-assisted NLP heuristics to accurately interpret requests like "CV ta one page koro" or "photo bad dao".
Challenge 03
Unifying multi-format resume ingestion across PDFs, Word documents, and scanned physical CVs.
Engineered Solution:
Built a unified parsing pipeline combining pdf-parse, mammoth, and Gemini Multimodal Vision to extract up to 25,000 characters without losing career milestones.
Challenge 04
Eliminating awkward whitespace and pagination overflow on standard A4 outputs.
Engineered Solution:
Engineered a smart vertical balancing algorithm that dynamically modulates section margins, line heights, and font scaling without altering candidate facts.
Challenge 05
Preserving guest progress and enabling friction-free migration upon account registration.
Engineered Solution:
Implemented HMAC-SHA256 signed guest sessions and an atomic database claim procedure (claimGuestResumes) that transfers all draft CVs and version history to the newly registered account.
Challenge 06
Rendering consistent, pixel-perfect A4 PDFs on Linux production containers without layout shifts.
Engineered Solution:
Configured headless Chromium with Playwright inside a Debian Docker container with pre-installed typography libraries and zero-margin print stylesheets.
13 / Result & Verification
Project Delivery & Technical Achievements
Successfully engineered and deployed Resumate AI to production on Render, delivering a high-performance conversational resume builder tested across multi-turn multilingual scenarios.
- Live production deployment on Render Cloud
- 8 verified regional CV templates with A4 PDF export
- Open-source codebase maintained on GitHub