Home / Projects / HireFit
HR Technology · Artificial Intelligence · Full-Stack

HireFit — Production AI CV Optimization & Career Matching Engine

A full-stack career platform that analyzes and optimizes resumes against target job descriptions with a strict Zero-Fabrication Guarantee, 6-category ATS scoring, document parsing, and interactive before/after diff editing.

Primary Role Full-Stack AI Solutions Architect
AI Integration Gemini API / LLM Structured Extraction
Document Formats PDF & DOCX (pdf-parse / mammoth)
ATS Export 3 Modular Print-Ready Templates
01

Project Overview

HireFit is an advanced, production-grade career technology platform built to address one of the most persistent bottlenecks in modern hiring: the disconnect between candidate resumes and specific employer expectations.

Rather than acting as a generic resume generator that creates synthetic, hallucinated career claims, HireFit is built on a Strict Zero-Fabrication Guarantee. It preserves the candidate's authentic employment history, education, and dates while parsing unstructured documents into structured schemas, scoring alignment across 6 evidence-based dimensions, and guiding candidates to reframe their real achievements in high-impact language aligned with Applicant Tracking Systems (ATS).

HireFit Production AI CV Optimization Platform
Figure 3.1: HireFit brand identity and verified platform visual assets.
02

The Real-World Problem

Job applicants and modern talent platforms encounter major friction when handling resumes:

✕ The Job Search Pitfalls

  • The Generic CV Trap: Sending the exact same CV to dozens of distinct roles without aligning keywords or highlighting relevant competencies.
  • AI Hallucination & Fabrication: Generic AI tools often invent job titles, companies, certifications, or statistics that candidates cannot defend during interviews.
  • Document Structure Destruction: Traditional resume optimizers discard the candidate's original layout, section ordering, and visual formatting.
  • Opaque Rejection Criteria: Candidates receive automated ATS rejections without knowing whether the gap was due to missing skills, poor phrasing, or formatting errors.

✓ The HireFit Engineering Approach

  • Zero-Fabrication Architecture: Hard-coded prompt constraints that strictly forbid hallucinating employers, credentials, or metrics.
  • 6-Category Multi-Dimensional Match Engine: Algorithmic scoring of Skills, Experience, Keywords, Responsibilities, Education, and ATS Readiness.
  • Structure-Preserving JSON Schema: Extracted resumes retain their precise section hierarchies, chronological orders, and formatting integrity.
  • Interactive Before/After Diff Inspector: Visual review pane enabling users to accept, reject, or fine-tune individual AI suggestions before committing.
03

The Solution Developed

HireFit combines robust MERN-stack engineering with modern large language model integration to provide a comprehensive, transparent career optimization platform:

📁

CV Document Library

Upload and manage multiple resume versions (.pdf, .docx up to 10MB) with Cloudinary cloud storage, local fallbacks, and safe cascade deletion.

⚙️

Multi-Format Text Extraction

Native parsing pipeline utilizing pdf-parse and mammoth with scanned PDF detection and clear user guidance.

🎯

Evidence-Based Match Scoring

Categorizes requirements as Matched, Partial Match, or Not Evidenced—avoiding negative assumptions while highlighting concrete gaps.

📄

ATS-Conscious PDF Export

Three professionally engineered templates (Modern Professional, Minimal ATS, Executive) with client-side print-ready PDF compilation.

04

Core Implemented Capabilities

  • Interactive CV Diff Editor: Visual side-by-side comparison showing proposed changes, the specific AI rationale, and granular controls to accept, decline, or manually adjust each edit.
  • Role-Specific Cover Letter Studio: Generates tailored, concise cover letters derived exclusively from verified CV experience with selectable tones (Confident, Professional, Technical).
  • AI Interview Coach & STAR Blueprints: Synthesizes tailored interview questions categorized into Technical, Behavioral, and CV-Specific questions, complete with STAR (Situation, Task, Action, Result) talking-point frameworks.
  • Integrated Application Tracker: Kanban-style application pipeline tracking submissions across Saved, Applied, Interview, Offer, and Rejected statuses linked to tailored CV snapshots.
  • Granular Role-Based Access: User accounts for candidates and administrative portals with system-wide telemetry on parsed CVs and average match scores.
05

System Architecture & Data Pipeline

The application couples a modular Express backend with LLM orchestration pipelines:

HireFit Full-Stack AI Architecture React + Node + LLM Pipeline
Client App
React.js SPA with Pure CSS styling, custom diff comparison components, HTML2PDF document generator, and reactive form state managers.
Document Engine
Multer file stream $\rightarrow$ MIME validator $\rightarrow$ pdf-parse / mammoth extraction $\rightarrow$ Cloudinary raw document bucket $\rightarrow$ text normalization pipeline.
AI Orchestration
Centralized LLM adapter (Gemini API / OpenAI SDK compatible) with JSON-schema output enforcement, prompt injection safeguards, and rule-based fallback handlers.
Database Layer
MongoDB with Mongoose. Schemas: User, Resume, JobAnalysis, OptimizationSnapshot, ApplicationTracker.
06

My Technical Contributions

  • AI Prompt Engineering & Guardrails: Authored zero-fabrication system prompts and structured output schemas ensuring the AI never invents phantom experiences.
  • Document Processing Pipeline: Implemented resilient text extraction supporting complex multi-column PDFs, tables, and Word documents up to 10MB.
  • Interactive Diff & Editor: Engineered the side-by-side diff UI allowing users to preview suggested phrasing improvements before writing back to the document model.
  • Export & Print Stylesheet: Created print-optimized CSS (@media print) and integrated client-side PDF generation for 3 ATS-conscious resume formats.
  • Backend & Database Architecture: Designed MongoDB models, JWT authentication, and Cloudinary cascade file asset cleanup.
07

Technology Stack

React.js Node.js Express.js MongoDB Atlas Gemini API / LLM pdf-parse mammoth (.docx) Cloudinary API HTML2PDF.js JWT & Bcryptjs Pure CSS3
08

Engineering Decisions & Ethical AI Design

1. The Zero-Fabrication Directive Ethical Engineering

Many AI resume tools hallucinate credentials to inflate match scores. When candidates are interviewed, these false claims collapse. In HireFit, I designed the system prompt to explicitly restrict the AI from introducing unverified companies, titles, degrees, or quantitative results. The AI is limited to restructuring real bullet points with stronger action verbs, highlighting genuine keywords, and clarifying context.

2. 'Not Evidenced' vs. 'Missing Skill' Taxonomy UX Decision

Traditional match engines tell candidates "You lack skill X". However, candidates often possess skills that simply were not written down. HireFit frames these as "Not Evidenced in CV", guiding candidates to add the experience if they possess it, or focus their cover letter on how they plan to acquire it.

3. Hybrid Cloud & Disk Storage Resilience Reliability

To guarantee zero downtime even during third-party cloud API outages, document uploads utilize a dual-path pipeline: uploading raw assets to Cloudinary while maintaining a secure disk fallback with automatic retry queues.

09

Impact & Operational Value

HireFit transforms the resume tailoring experience from an agonizing, hours-long chore into an empowering, evidence-based process completed in minutes. By maintaining honesty and factual accuracy at its core, it gives job seekers the confidence that their application accurately reflects their abilities while maximizing technical visibility in ATS screening systems.

6
Scoring Dimensions
0
Fabricated Facts Allowed
3
ATS PDF Templates
10MB
Max File Upload Capacity