Offerum

All-in-one AI career platform for resumes, ATS scoring, cover letters, and interviews.

Comment

💡 Inspiration

In today's brutal job market, sending out resumes often feels like shouting into a void. Over 75% of resumes are automatically filtered out by Applicant Tracking Systems (ATS) before a human recruiter ever sees them. Job seekers spend countless grueling hours manually tailoring resumes, writing repetitive cover letters, and guessing what keywords algorithms want.

We asked ourselves: Why should qualified candidates lose opportunities simply because they didn't format a bullet point or match an arbitrary parsing algorithm?

This inspired us to build Offerum—an intelligent, all-in-one AI career copilot designed to level the playing field, beat the black-box ATS filters, and empower candidates from initial application to final interview.

🚀 What it does

Offerum (offerum.work) is a unified career acceleration platform that turns the chaotic job search into a structured, high-conversion workflow:

  • 📄 Smart AI Resume Builder: Offers modern, recruiter-approved, and ATS-tested templates that ensure formatting never breaks during machine parsing.
  • 🎯 Intelligent ATS Compatibility Checker: Scans the user's resume against target Job Descriptions (JD), pinpointing missing keywords, structural red flags, and semantic relevance.
  • ✉️ Role-Specific AI Cover Letter Generator: Crafts persuasive, tailored cover letters aligned with the role, target company culture, and user experience in seconds.
  • 🎙️ AI Mock Interview Coach: Simulates realistic, role-specific technical and behavioral interviews with instant actionable feedback on clarity, structure, and depth.
  • 📊 Kanban Job Application Tracker: Centralizes applications, interview schedules, notes, and deadlines into an intuitive CRM for your job hunt.
  • 🧩 Browser Extension Integration: Seamlessly captures job postings from LinkedIn, Indeed, and company career portals with a single click.

🛠️ How we built it

Offerum is engineered for speed, responsiveness, and deterministic precision:

  • Frontend & UI/UX: Built with React 18, TypeScript, Vite, and Tailwind CSS, styled using accessible Radix UI / shadcn/ui primitives and animated with smooth micro-interactions.
  • Backend & Database: Powered by Supabase (PostgreSQL), leveraging Row Level Security (RLS) to protect sensitive personal and career data, with Supabase Storage managing resume assets.
  • AI & NLP Pipeline: Combines LLM prompt engineering with heuristic text parsing to analyze resume structure and compute semantic alignment.

📐 The ATS Scoring Formula

To provide objective, actionable feedback, we evaluate candidate resumes using a multi-factor scoring function:

$$S_{\text{ATS}} = w_1 \cdot K_{\text{exact}} + w_2 \cdot S_{\text{semantic}} + w_3 \cdot F_{\text{structure}}$$

Where:

  • $K_{\text{exact}} \in [0, 1]$ represents the exact match ratio of mandatory skills and keywords extracted from the JD.
  • $S_{\text{semantic}} \in [0, 1]$ calculates the cosine similarity of domain concepts and experience embeddings: $$S_{\text{semantic}} = \frac{\mathbf{v}{\text{resume}} \cdot \mathbf{v}{\text{JD}}}{|\mathbf{v}{\text{resume}}| |\mathbf{v}{\text{JD}}|}$$
  • $F_{\text{structure}} \in [0, 1]$ assesses formatting compliance (parsable contact details, standard section headings, chronological consistency).
  • $w_1, w_2, w_3$ are balanced weights calibrated against real-world ATS parsing engines.

⚡ Challenges we ran into

  1. Aesthetic Design vs. ATS Parsability: Highly graphical resumes look great to humans but break conventional ATS parsers. We had to carefully engineer the DOM structure, CSS print media stylesheets, and PDF rendering pipeline to guarantee $100\%$ text-layer extractability while maintaining modern visuals.
  2. Preventing AI Hallucinations in Resumes: Resumes demand absolute factual fidelity. We implemented strict prompt constraints and structured JSON schema validations to ensure the AI enhances metrics and phrasing without inventing false credentials.
  3. Multi-Locale Career Conventions: Resume norms vary significantly between regions (e.g., standard US 1-page resumes vs. Asian/European CV conventions). We built a modular internationalization (i18n) engine and localized layout templates.

🏆 Accomplishments that we're proud of

  • High ATS Parsing Rate: Tested against common parsing tools with an exceptional parsability rate across all templates.
  • Lightning-Fast Generation: Reduced the time required to tailor an application to a new role from 45 minutes to under 60 seconds.
  • Comprehensive Global Architecture: Successfully created an extensible localized template generator supporting multiple languages and industries.

📚 What we learned

  • Deconstructing the ATS Black Box: We gained deep technical appreciation for how modern enterprise HR systems parse headings, dates, and skill hierarchies.
  • Prompt Chaining for Career Coaching: Single-shot LLM prompts often provide generic advice; dividing the interview and resume review into multi-turn evaluation stages yields drastically more constructive feedback.

🔮 What's next for Offerum

  • Autonomous Auto-Apply Agent: Enhancing the Chrome Extension to safely autofill complex multi-step application portals (Workday, Greenhouse, Lever).
  • Voice-Enabled Real-Time Interviewing: Upgrading the mock interview experience with real-time voice latency and speech cadence analysis.
  • Post-Interview Debrief Analysis: Allowing candidates to paste actual recruiter questions to generate personalized follow-up notes and counter-offers.

Built With

Share this project:

Updates