SkillDecayTracker

SkillDecayTracker monitors the demand of your professional skills, highlights which are losing relevance, and provides data-driven guidance to upskill or pivot effectively.

Comment 1

SkillDecayTracker

Inspiration

The professional world moves fast, and skills can quickly lose relevance. I noticed peers — and myself — spending time learning technologies that were fading in demand. This inspired SkillDecayTracker, a tool to track skill trends and guide upskilling decisions.

What it does

SkillDecayTracker analyzes trends from job postings, GitHub, and Stack Overflow to highlight:

  • Skills losing relevance 📉
  • Skills growing in demand 📈
  • Adjacent skills to pivot into 🔄
  • Suggested learning resources and micro-projects 📚 Users input their skills and career goals, and the platform provides actionable insights to prioritize learning.

How we built it

  1. Frontend: Next.js + TailwindCSS for a responsive dashboard
  2. Backend: FastAPI serving trend and skill data
  3. Data: Pre-collected datasets, GitHub topics, Stack Overflow trends
  4. NLP: spaCy + keyword extraction for skill mapping
  5. Visualizations: Chart.js for trend lines and risk bands
  6. Deployment: Vercel (frontend) + Render (backend) + Supabase for data

Skill Decay Model

We compute the trend slope ( m ) using linear regression: m=n∑(xy)−∑x∑yn∑(x2)−(∑x)2m = \frac{n\sum(xy) - \sum x \sum y}{n\sum(x^2) - (\sum x)^2}m=n∑(x2)−(∑x)2n∑(xy)−∑x∑y​ Where ( x ) represents time periods and ( y ) represents skill demand metrics. python# Example: compute skill trend slope import numpy as np slope = np.polyfit([1,2,3,4], [50,45,40,30], 1)[0] print("Trend slope:", slope) Risk Classification: Skills with ( m < -5 ) are flagged as high-risk for decay.

Challenges we ran into

  • Limited access to real-time labor market data → used static datasets for MVP
  • Normalizing skill mentions across multiple platforms
  • Presenting trends clearly without overwhelming the user
  • Avoiding overpromising skill predictions → used risk bands instead of exact "half-life"

Accomplishments that we're proud of

  • Built a working dashboard showing skill decay trends
  • Developed adjacent skill recommendations using cosine similarity: ( \text{similarity}(A,B) = \frac{A \cdot B}{||A|| \cdot ||B||} )
  • Designed a simple, intuitive visualization for fast decision-making
  • Laid the foundation for a data-driven career guidance tool

What we learned

How to combine multiple data sources into actionable insights Translating abstract skill trends into user-friendly visualizations Ethical AI: providing guidance without overpromising The importance of transparency — showing confidence intervals rather than false precision

What's next for SkillDecayTracker

Integrate more real-time labor market data (LinkedIn API, Indeed scraping) Add personalized alerts for at-risk skills Recommend micro-projects and courses automatically Expand to team and enterprise skill tracking Implement collaborative filtering: ( \hat{r}{ui} = \bar{r}u + \frac{\sum{v \in N(u)} \text{sim}(u,v)(r{vi} - \bar{r}v)}{\sum{v \in N(u)} |\text{sim}(u,v)|} )

Built With

  • chart.js
  • fastapi(python)
  • githubapi
  • next.js
  • react
  • spacy
  • supabase(postgresql+auth)
  • tailwindcss
Share this project:

Updates