Capstone 5 — DevPortfolio

Tagline: A developer portfolio + business-card app powered by CoreML (on-device resume parsing) and ARKit (drop your projects as 3D cards in the world), structured with The Composable Architecture, and shipped end-to-end through Apple Review.

Tech stack: CoreML · Vision · ARKit · RealityKit · TCA (The Composable Architecture) · App Store submission

Time budget: ~2.5 weeks

What this capstone proves:

  • You can ship CoreML with a real, useful model (text classification — tag a resume snippet as “experience,” “education,” “skill”)
  • You can build a working ARKit experience (3D project cards placed in the real world)
  • You can architect a non-trivial app with TCA, demonstrating you understand redux-style unidirectional flow in a Swift context
  • You can walk a reviewer through the full App Store submission including metadata, screenshots, privacy nutrition label, App Review Notes, and post-submission iteration

The 30-second pitch

“DevPortfolio is a developer’s interactive business card — you upload your resume, an on-device CoreML model classifies each paragraph as experience / skill / education / project, and the app builds a structured portfolio you can share via a QR code or a deep link. The AR mode lets you drop your projects into the room as floating 3D cards a recruiter can walk around. The whole app is structured in TCA, which gave me a single deterministic state tree and made testing the resume parser end-to-end straightforward — every classifier output flows through a Reducer I can assert against.”

Why this capstone

Three independent senior-signal technologies in one app: ML, AR, and a non-default architecture. None of them are toy demos — the ML model does real classification on real text, the AR mode interacts with detected planes, and TCA structures the actual flow, not just the README. Combined with a full App Store submission walkthrough, this is the capstone that shows “I can ship anything.”

Files

  1. requirements.md — feature list, user stories, acceptance criteria
  2. architecture.md — TCA structure, ML pipeline, AR session lifecycle, ADRs
  3. implementation-guide.md — step-by-step build walkthrough
  4. hardening-checklist.md — production + Review + privacy
  5. interview-talking-points.md — pitch and Q&A

What “done” looks like

  • App live on the App Store
  • Resume parsing works on three test resumes; classifier accuracy ≥ 80%
  • AR mode places at least 3 cards on a detected horizontal plane
  • TCA structure visible — root AppFeature, child features for ResumeParser, Portfolio, ARSession
  • Full App Review walkthrough documented in the README

Next: Requirements