SkillScan — AI Career Intelligence for Students

🌟 Inspiration

Every semester I watched brilliant friends get rejected from jobs they were perfect for.

Not because they lacked talent. Not because they didn't work hard. But because nobody ever told them what was actually missing from their profile.

A friend with a Computer Science degree applied to 47 data science roles in 3 months. 47 rejections. Zero feedback. Zero clarity. She had no idea SQL Advanced was blocking every single application.

That silence — that information gap between students and employers — is what SkillScan was built to destroy.

"The gap between a student and their dream job is not talent. It's information."

💡 What It Does

SkillScan is a 7-feature AI career intelligence platform that gives students everything they need to go from graduate to hired in 30 days.

📄 Resume Scanner

Upload resume → Gemini AI scores it 0–100 → flags every weak section → gives exact rewrites. Not vague advice. Specific fixes that make ATS systems and recruiters stop scrolling.

🔍 AI Career Gap Analyzer

Select dream role → AI compares your skills against real market demand → Gap Score with every missing skill ranked by hiring priority. Learn the right skills in the right order.

🗺️ Skill Roadmap Generator

Gap analysis feeds into a personalized 30-day learning plan — daily tasks, specific resources, milestones, and an estimated job-ready date. All generated in seconds.

🎯 Job Match Score

Paste any JD from LinkedIn or Naukri → instant match percentage → green skills you already have → red skills blocking this specific job. Know before you apply.

🎤 AI Interview Simulator

Live AI mock interview → 8 realistic questions → scored across Relevance, Clarity, and Technical accuracy → full post-interview report with ideal answer hints per question.

💰 Salary Intelligence Engine

See current market value vs projected salary after closing gaps. Skill-by-skill salary impact. City-wise comparison across Bangalore, Mumbai, Hyderabad. AI-generated negotiation script.

🔗 LinkedIn Profile Analyzer

AI scores every LinkedIn section 0–10 → rewrites headline and About section → keyword density analysis → recruiter click probability before and after fixes.

🛠️ How We Built It

Built as a solo developer using Lovable.dev for rapid full-stack development — combining React, TypeScript, Tailwind CSS, and Supabase into a production-ready platform in record time.

Frontend Layer: React 18 + TypeScript + Tailwind CSS Framer Motion (animations) Recharts (data visualizations) Web Speech API (voice interview answers) Deployed: Lovable Cloud → Netlify CDN

AI Intelligence Layer: Google Gemini Flash API 7 specialized system prompts:

Resume weakness detection Skill gap classification + priority ranking 30-day roadmap constraint scheduling Interview question generation Answer evaluation (3 dimensions) Salary range calculation per skill/city LinkedIn section scoring + rewriting

Data Layer: Supabase (PostgreSQL + Realtime) Row Level Security enforced Tables:

student_skill_profiles resume_scan_history roadmap_progress interview_sessions salary_intelligence_cache linkedin_analysis_history job_match_history

The Gap Score Algorithm:

$$\text{Career Readiness} = \frac{\sum_{i=1}^{n} w_i \cdot p_i}{\sum_{i=1}^{n} w_i} \times 100$$

Where:

$ w_i $ = hiring frequency weight per skill
$ p_i $ = student proficiency (0, 0.5, or 1)

Priority Ranking Formula:

$$\text{Priority}_i = f_i \times d_i$$

Where:

$ f_i $ = frequency in real job postings
$ d_i $ = difficulty-to-acquire inverse weight

High frequency + lower difficulty = learn first.

⚔️ Challenges We Ran Into

1. Resume Parsing Across Different Formats

PDFs from different colleges use completely different layouts — two columns, tables, creative designs. Early builds misread these consistently. Fixed by preprocessing through Gemini Vision before skill extraction — handles any resume format regardless of structure.

2. Gap Score Gaming

Beta testers found they could inflate their score by selecting all skills as Advanced. Fixed with a Gemini verification layer — AI asks 1–2 quick questions per claimed skill and adjusts proficiency based on actual answers. You cannot fake your way to a high match score.

3. Interview Evaluation Latency

First builds had 5–7 second lag between answer submission and feedback appearing. Fixed by streaming the Gemini evaluation response and rendering feedback components progressively as tokens arrive — reducing perceived wait from 7 seconds to under 1.5.

4. Salary Data Granularity

Generic salary ranges like ₹3–15 LPA are meaningless. Fixed by building a skill-combination prompt asking Gemini to calculate salary impact per specific skill at specific proficiency level in specific city — returning granular data instead of broad ranges.

5. LinkedIn Semantic Keyword Matching

"Proficiency in Python" vs "Python experience" vs "Python 3.x" all mean the same thing but look different to basic parsers. Fixed with semantic matching via Gemini — understands meaning and intent, not just exact strings.

🏆 Accomplishments That We're Proud Of

🎯 7 fully working AI features — each solving a distinct piece of the student hiring problem, all built and deployed by a solo developer.

🎤 Voice-powered interview simulator — students can speak answers using Web Speech API and get scored in real time. No other career tool does this.

💰 Skill-level salary intelligence — not generic ranges but specific LPA impact per skill per city, making the ROI of learning instantly visible.

🔗 LinkedIn rewrite engine — AI doesn't just score your profile, it rewrites your headline and About section word-for-word, ready to copy.

⚡ Under 2-second AI response — all 7 Gemini integrations optimized for speed with streaming responses and progressive UI rendering.

📊 Mathematical gap scoring — weighted multi-dimensional algorithm inspired by the Maslach Burnout Inventory adapted for career readiness — not a simple checklist.

🌍 Impact at scale:

$$\text{Students Helped} = 8{,}000{,}000 \times 0.01 = 80{,}000$$

$$\text{Additional Hires} = 80{,}000 \times 0.30 = 24{,}000$$

Every hire = one career launched, one family changed.

📚 What We Learned

Information asymmetry IS the skills gap. Most students don't fail because they're unqualified — they fail because they don't know what qualified looks like for their specific target role in the current market.

AI tone and specificity matter equally. A technically accurate but vague response is useless. SkillScan works because Gemini returns specific, actionable output — not general career advice.

Overwhelming users destroys retention. Early versions showed 20 missing skills at once. Every user felt hopeless and closed the app. The fix: show top 3 priority gaps only. Focus beats comprehensiveness every time.

Speed is a feature, not a bonus. Every second of AI wait time costs engagement. Streaming responses transformed how the interview simulator felt — from slow to magical.

Solo building forces brutal prioritization. Every feature had to earn its place. No feature survived that couldn't be explained in one sentence and demoed in 30 seconds.