SkillTwin

Know what you know. Fix what you don't. SkillTwin turns scattered skill signals into one evidence-backed readiness score — and closes the gap with real missions, not more courses.

Comment

Challenge 5: Career Readiness

Inspiration

We kept hearing the exact same story from final-year engineering students: "I’ve solved 350 DSA problems, built three projects, and completed two certifications — but I still got rejected in the technical round without knowing why."

The problem is fragmented signals:

  • Coding platforms track LeetCode count, not engineering depth.
  • Course platforms track video completion, not demonstrated ability.
  • Resumes show self-proclaimed buzzwords that recruiters distrust.
  • GitHub repositories sit unverified.

None of these systems communicate with each other, and none answer the single question that actually matters: "Am I truly ready for the exact engineering role I want?" We built SkillTwin to diagnose, prioritize, and close competency deficits before the interview, turning speculative resumes into evidence-verified digital career twins.

What it does

SkillTwin creates a dynamic Digital Skill Twin of a student and continuously benchmarks it against real industry requirements for their target engineering career:

  • Evidence-Backed Skill Graph: Ingests academic milestones, DSA problem counts, project repositories, languages, and certifications into a scored competency graph \( (0 - 100\%) \).
  • Deterministic Priority Formula: Ranks skill deficits using a transparent mathematical formula: $$ \text{Priority Score} = \text{Skill Gap} \times \text{Industry Importance Weight} \times \text{Career Relevance Weight} $$ Ensuring students focus on mission-critical deficits rather than over-optimizing skills they've already mastered.
  • Interactive Hands-On Evidence Missions: Generates targeted technical missions for the #1 critical gap (e.g., containerizing microservices, writing database indexing benchmarks, or configuring CI/CD test automation).
  • Direct In-Modal Code & Artifact Evaluator: Students can either submit GitHub repositories or use our in-browser syntax editor to inspect and test Dockerfile, compose.yml, and tests/ with instant rubric evaluation and artifact tree matching.
  • "Skill Defense" & Mock Interview Simulations: 60-second interactive scenario drills and mock architecture defenses that test real-world trade-offs (cache stampedes, row-level locks, zero-downtime rolling deploys) with immediate evidence bonus rewards \( (+5 \text{ to } +10 \text{ pts}) \).
  • Cross-Role & Predictive "What-If" Simulators:
    • Cross-Role Matrix: Benchmarks the twin across 4 roles simultaneously (Backend, Full-Stack, DevOps, Data Analyst).
    • Predictive Trajectory Engine: Interactive sliders project 30-day readiness growth curves and forecast exact campus interview-ready dates.
  • Verified Digital Credential Export: Generates shareable, printable credential cards with verifiable IDs (ST-2026-XXXXX), QR verification codes, and an audit trail for LinkedIn and recruiter applications.
  • Institutional Telemetry & Faculty Intervention:
    • College admins access institutional skill deficit heatmaps and 1-click cohort CSV exports.
    • Curriculum Intervention Generator: Automatically designs 2-week faculty lab syllabi tailored to the cohort's #1 gap.
    • Faculty Nudge Loop: Professors can assign priority missions directly to students with custom coaching notes.

How we built it

We built SkillTwin as a high-performance, zero-friction full-stack monorepo designed for 100% deterministic reliability:

  • Frontend: React 18, Vite, Tailwind CSS, Recharts (Radar, Bar, and Forecast Trajectory charts), Lucide Icons, and Canvas Confetti.
  • Backend & Core Engine: Node.js, Express, and a custom universal database engine implementing the Prisma ORM API for portable, zero-dependency persistence.
  • Scoring & Rubric Engine: Hand-authored industry weighting matrices, deterministic heuristic scoring, and a regex/file-tree pattern-matching rubric evaluator.
  • DevOps & Testing: Multi-stage Dockerfile, docker-compose.yml, GitHub Actions CI/CD automation matrix, and a comprehensive Node test suite (5/5 tests passing).
  • Persistent Judge Demo Toolbar: A floating toolbar enabling 1-click persona switching (Alex Chen [58%], Sarah Miller [86%], Jordan Lee [36%], Prof. Davis [Admin]) with zero setup friction.

Challenges we ran into

  • Avoiding Black-Box AI Hallucinations: Initial instincts leaned toward using LLMs to grade arbitrary code repositories. Recognizing that unpredictable model outputs ruin credibility and live demos, we deliberately engineered an automated, rule-based rubric matcher and file-tree inspector that produces deterministic, explainable results every single run.
  • Designing Fair Gap Weighting: Raw arithmetic percentage gaps penalized students unfairly on low-importance secondary skills. Developing our triple-weighted formula \( (\text{gap} \times \text{importance} \times \text{relevance}) \) produced natural, career-accurate prioritization.
  • Ethical Framing of Career Readiness: We ensured the Career Readiness Index is explicitly presented with a transparent disclaimer: "An evidence-backed readiness indicator based on demonstrated proof, not an employment guarantee."

Accomplishments that we're proud of

  • Complete End-to-End Feedback Loop: Seamless progression from profile evidence \( \to \) gap prioritization \( \to \) code artifact evaluation \( \to \) skill delta award \( \to \) updated readiness score.
  • Multi-Role & Predictive Simulators: Students can visualize cross-career applicability and forecast their exact interview-readiness trajectory with interactive sliders.
  • Two-Way Institutional Workflow: Bridging college administrators and students through automated curriculum syllabus generation and faculty mission nudges.
  • 100% Verifiable & Demo-Ready: Built with zero mock placeholders, featuring instant 1-click test templates, complete automated test suites, and containerized CI/CD.

What we learned

  • Students need clarity over content: Students don’t suffer from a lack of learning materials; they suffer from navigational ambiguity. Knowing the exact priority deficit to fix changes student behavior faster than another 40-hour course ever could.
  • Evidence beats claims: Verifiable artifacts (passing tests, Docker health checks, indexed queries) are infinitely more valuable to recruiters than unverified resume bullet points.
  • Scoping with discipline creates superior products: Building a robust, deterministic core with rich interactivity creates a far more convincing, pitch-ready platform than an ungrounded AI wrapper.

What's next for SkillTwin

  • AI-Assisted Deep Code Review: Introduce bounded LLM sidecars to provide nuanced semantic code reviews atop our deterministic rubric checks.
  • Live Job Market Ingestion: Dynamically calibrate role skill weights by ingesting live job postings from real-time API feeds.
  • Institutional College Pilots: Partner with engineering college placement cells to benchmark cohort skill twins against actual placement conversion rates.
  • Corporate Recruiter Talent Search: Launch the employer portal allowing hiring managers to search, filter, and shortlist pre-verified candidates by demonstrated skill evidence. ```

Built With

Share this project:

Updates

Submission history