Inspiration

SkillBridge was inspired by a harsh reality in today’s job market: the "ATS Black Hole." Many people have the skills and motivation, yet still struggle to get hired because they do not clearly understand what employers expect. We noticed that job roles often look the same on paper but vary greatly between companies. This confusion leads to unstructured learning, repeated rejections, and wasted effort. Furthermore, professional career coaching and mock interviews are prohibitively expensive for recent graduates. SkillBridge was created to democratize career coaching, closing the gap by giving job seekers clarity, direction, and confidence through the power of AI.


What it does

SkillBridge is an end-to-end, AI-powered employability platform that acts as a 24/7 personalized career coach.

  • Smart Parsing: Users upload their CVs (PDF or DOCX), and the system instantly extracts their professional profile.
  • Gap Analysis & Roadmaps: Users input their target job role. The system analyzes their skills against real-world job requirements, identifies specific skill gaps, explains their readiness level, and generates a personalized, step-by-step learning roadmap.
  • Smart Draft Management: Users can save their progress mid-analysis and resume later. The system intelligently auto-deletes drafts once an analysis is successfully completed to keep the dashboard clean.
  • Interactive AI Coach: A built-in Chat Widget allows users to ask career questions, featuring full conversation control—users can delete individual messages or clear the entire chat history to reset the AI's memory.
  • Live Voice Mock Interviews: Users can toggle "Live Voice Mode" to conduct real-time, low-latency mock interviews, practicing the STAR method out loud with the AI, featuring a fully scrollable and responsive interface for seamless accessibility.
  • Exportable Insights: Users can download their AI chat histories and analysis reports directly as beautifully formatted PDFs.
  • Sharing: Users can share their analysis report by generating either a temporary (1 hour, 8 hours, 1 day, or 1 week), or permanent link. (These links can be expired/deleted anytime on user's choice).

How we built it

We built SkillBridge using a modern, full-stack architecture optimized for speed and AI integration. The frontend provides a sleek, responsive user experience, while the backend manages secure user sessions and a credit-based usage system.

For the AI Engine, we orchestrated multiple Google Gemini models to balance cost and performance:

  • Gemini 3.1 Pro & 3.0 Flash: Used for deep semantic reasoning and balanced analysis tasks.
  • Gemini 3.1 Flash Lite: Used for rapid, cost-effective CV parsing and structured JSON generation for the Gap Analysis.
  • Gemini 2.5 Flash: Powers the contextual text-based Chat Widget.
  • Gemini Live API (Audio): Handles the real-time, bidirectional voice streaming for mock interviews.

We also integrated document processing libraries for parsing DOCX files and built a custom Markdown-to-PDF rendering engine to allow users to export their chats and roadmaps.


Challenges we ran into

  1. Unstructured Data: Handling the variety and ambiguity of language in CVs was difficult. The same skill is often described in many different ways, making simple keyword matching ineffective. We had to heavily engineer our Gemini prompts to ensure semantic matching and consistent JSON outputs.
  2. Real-Time Audio Streaming: Integrating the Gemini Live API required complex state management. We had to manually handle Web Audio API contexts, PCM16 audio encoding/decoding, and buffer management to ensure the voice mock interviews felt natural and interruption-friendly.
  3. Custom PDF Rendering: Converting dynamic AI Markdown (with bolding, lists, and code blocks) into a clean PDF required writing a custom word-wrap and pagination engine from scratch.
  4. Complex State & Memory Management: Giving users granular control over the AI Chat Widget (like deleting specific messages or clearing history to reset context) required careful synchronization between the frontend UI state and the AI's context array. Additionally, making the Voice Mode interface fully responsive and scrollable without disrupting the active audio session was a tricky UI challenge.

Accomplishments that we're proud of

We are incredibly proud of building a system that goes far beyond basic keyword matching. By integrating the Gemini Live API, we successfully created a voice-native mock interview experience that genuinely feels like talking to a human recruiter. We are also proud of our dynamic UI, the seamless Firebase credit system, and our custom PDF export engine. Most importantly, SkillBridge focuses on employability outcomes rather than just displaying data, making it a highly practical and impactful tool.


What we learned

Through this project, we learned that employability is not just about skills, but about alignment and clarity. Technically, we gained massive experience in designing explainable AI systems, orchestrating multiple LLMs based on task complexity (Pro vs. Flash vs. Flash Lite), working with WebSockets for live audio, and creating user-focused solutions that address real-world anxiety around job hunting.


What's next for SkillBridge – Bridging Skills to Decent Work

Since we successfully implemented interview readiness via our Live Voice feature, our next steps focus on expansion:

  • LinkedIn Integration: Allowing users to import their profiles directly via URL.
  • Automated Cover Letters: Generating highly tailored cover letters based on the Gap Analysis data.
  • ATS Resume Rewriter: A tool that automatically suggests rewrites for specific bullet points to bypass Applicant Tracking Systems.
  • SDG 8 Alignment: Our long-term vision is to make SkillBridge a continuous career companion that helps individuals adapt to changing job markets and move toward stable, decent work, in line with UN Sustainable Development Goal 8.

Built With

  • cloud-firestore
  • css3
  • express.js
  • firebase-authentication
  • framer-motion
  • gemini-2.5-flash
  • gemini-3.1-flash-lite
  • gemini-live-api-(audio)
  • google-genai-sdk
  • html5
  • javascript
  • jspdf-(pdf-generation)
  • lucide
  • mammoth.js-(docx-parsing)
  • node.js
  • pdfjs-dist-(pdf-parsing)
  • react-19
  • tailwind-css
  • typescript
  • vite
Share this project:

Updates