Case Study 01
SkillVision AI
Intelligent Career & Skill Intelligence System
SkillVision is a full-stack AI career copilot and placement platform. The verified build is React 19 + Vite + Tailwind, an Express + TypeScript backend, Google Gemini 2.5 Flash agents, JWT + bcrypt auth, and a file-backed JSON store — deployed at skillvisionai.onrender.com. There is no vector database in this build.
01 · Overview
Overview
- AI career copilot: upload a resume, get an ATS score (0–100), see skill gaps on a 6-axis radar, receive a learning roadmap, matched jobs and mock + technical interview coaching.
- Solves fragmented prep — resume checking, planning, job search and interviews in four tools — with one dashboard and one auth session.
- For students and placement seekers. Technically notable for 13+ agents centralised in server/agents.ts behind REST, with a local JSON file as the database.
02 · The Problem
The Problem
- Career prep is scattered across ATS checkers, roadmap blogs, job boards and mock interviews that share no context.
- Generic advice ignores the candidate's resume and target role.
- No loop connects resume evidence → gaps → tasks → jobs → interview readiness.
03 · The Solution
The Solution
- One workspace where the resume is the source of truth for downstream agents.
- Modular agents (resume, quiz, recommendations, skill-gap, roadmap, courses, jobs, custom-job, interview, tech-interview, tips) over REST.
- Dashboard summary + tabs: dashboard, resume, assessment, roadmap, jobs, interview, tech_interview, progress.
04 · How It Works
How It Works
01 · 1. Upload resume-
POST /api/upload-resume; analysis returns skills, ATS score, tailoring tips.
02 · 2. See gaps+
/api/skill-gap + 6-axis radar.
03 · 3. Get roadmap+
/api/roadmap with task toggles + /api/courses.
04 · 4. Match jobs+
/api/jobs + custom analysis with cover-letter pitch.
05 · 5. Practice+
interview agents (STAR, coding studio) evaluate answers.
06 · 6. Track streaks+
DailyStreakCounter + dashboard summary.
Select a stage to expand it. Every stage maps to code in the repository.
05 · System Architecture
System Architecture
React 19 + Vite 6 + Tailwind 4 tab app; SkillRadarChart, JobMatchingHub, DailyStreakCounter, GoogleAuthDialog.
Express 4 + TS (port 3000); serves dist/ in prod; /api auth, resume, roadmap, courses, jobs, interview.
@google/genai Gemini 2.5 Flash via server/agents.ts (13+ agents).
server/db.ts file JSON (users, profiles, resumes, assessments, roadmaps, jobs, sessions, streaks). No vector DB.
06 · Core Features
Core Features
ATS Resume Analyzer (0–100)
ATS Resume Analyzer (0–100) — grounded recommendations. Scope note: full resume builder not proven; only tailoring tips + cover-letter pitch.
6-Axis Skill Radar
6-Axis Skill Radar — visual gap analysis via /api/skill-gap.
Dynamic Roadmaps
Dynamic Roadmaps — generated paths with toggleable tasks + courses.
Job Matching Hub
Job Matching Hub — fit feedback + custom analysis.
AI Mock + Tech Interview Coach
AI Mock + Tech Interview Coach — questions, evaluation, STAR, coding studio.
Streaks + Dashboard
Streaks + Dashboard — momentum in one view.
07 · Technical Implementation
Technical Implementation
- React 19, Vite 6, Tailwind 4, recharts, lucide-react, motion; SEO wired to Render URL.
- Node + Express 4 + TS, tsx/esbuild, dotenv, JWT/bcrypt + Google OAuth endpoint.
- Gemini 2.5 Flash only; no embeddings/vector libs.
- npm run build + npm start serves dist/ from Express on Render.
08 · My Contribution
My Contribution
- Full-stack build: tab UI, REST surface, Gemini wiring, auth, JSON model, dashboard aggregation.
- SEO + deployment to Render with sitemap/robots/canonical.
- README with dev and build flows.
09 · Tech Stack
Tech Stack
- Frontend
- React 19 · Vite 6 · Tailwind 4.
- Backend
- Node · Express 4 · TypeScript.
- AI
- Gemini 2.5 Flash, 13+ agents.
- Auth
- JWT · bcrypt · Google OAuth.
- DB
- file JSON (no vector DB).
- Deploy
- Render.
10 · Challenges & Decisions
Challenges & Decisions
- File JSON limits concurrency → narrow schema + dashboard aggregation → single-file persistence on Render.
- Many agents, one model → centralised prompts in agents.ts → consistent behaviour.
- Similar-named RAG prototype exists → describe only this deploy → no vector claims.
11 · Result
Result
- Working: auth, ATS analysis, radar, roadmaps, courses, job matching, interviews, streaks, dashboard.
- Live on Render with SEO docs.
- Limits: no vector retrieval, no standalone builder.
12 · Links