00
PORTFOLIO.INIT
Skip to content
Mobile & AI Engineer

Hi! I am Ryan

Full-stack engineer building production apps with React Native, Swift, computer vision, and privacy-first AI.

View Work Contact Me
Shipped Apps 4 released
HabitForge HabitForge Seekr Seekr TrainTogether TrainTogether DigiMimic DigiMimic
All live on the App Store
Real products at the intersection of mobile, AI, and native hardware

Product-Focused
Engineer.

I build production-minded mobile products that combine applied AI, native device capabilities, and careful UX. Across HabitForge, Seekr, TrainTogether, and DigiMimic, plus this portfolio website, I have shipped work spanning React Native, Swift, Gemini-backed workflows, on-device models, computational photography, LiDAR scanning, spatial search, and resilient frontend systems.

At the University of Toronto, I am strengthening my foundations in algorithms, math, and software design while continuing to ship real products. My volunteer work in tech support and teaching sharpened my communication and collaboration, and I care about building software that is technically rigorous, reliable, and genuinely useful.

0 Years Exp
0 Projects
0 Tech Stacks
Shipped apps spanning habits, spatial computing, fitness, fintech, and computational photography

Featured
Projects

HabitForge Preview

HabitForge

React Native Expo TypeScript Swift Gemini WidgetKit
  • Structured around habit, finance, and rewards domains so the product scales without feature coupling
  • Natural-language expense capture and budget workflows that reduce friction for repeated daily logging
  • Opt-in Gemini insights run from a user-supplied API key, keeping AI usage transparent and fully client-side
  • Native widget bridge with a consolidated sync pipeline for habits, budgets, and expenses
  • Gamification, finance visibility, and AI guidance in one cohesive consumer product loop
Seekr Preview

Seekr

Swift RoomPlan ARKit LiDAR CloudKit SceneKit Multipeer
  • RoomPlan and LiDAR turn a room walkthrough into an editable 3D model with nested Furniture, Container, and Item tags
  • AR wayfinding relocalizes a saved world map to guide you back to a tagged item, and RoomPlan's StructureBuilder merges rooms into a whole-home 3D twin
  • A prebuilt, relevance-ranked search index plus on-device barcode and receipt OCR keep large inventories fast to search and fill in
  • Optional CloudKit backup and household sharing sit alongside a fully offline MultipeerConnectivity sync path, so cloud use stays opt-in
  • PDF, CSV, and insurance-claim exports plus a portable JSON archive cover backup, claims, and moving data between devices
TrainTogether Preview

TrainTogether

Swift SwiftUI MediaPipe On-device LLM (Qwen) GRDB / SQLite HealthKit StoreKit
  • Real-time MediaPipe pose pipeline across 30+ exercises, each with its own rep counter and form evaluator sharing one scoring framework
  • On-device LLM coaching (Qwen via llama.cpp) that runs fully offline after a one-time model download, with a rule-based fallback in Low Power Mode
  • Nutrition and hydration tracking with optional, consent-gated AI food-photo scanning proxied to keep the vision API key off-device
  • HealthKit sync plus a Home Screen/Lock Screen widget and Live Activities for daily workout and nutrition progress
  • Free core workout experience with a Pro subscription that raises AI coaching and food-scan limits
IBKR Portfolio Analyzer Preview

IBKR Portfolio Analyzer

Python FastAPI Next.js TypeScript Docker Gemini ib-insync
  • IBKR Gateway via ib-insync with multi-account USD/CAD consolidation, live P&L tracking, and connection diagnostics
  • Deterministic stock-scoring engine (GBDT-style decision trees) with suitability checks, IPS drift detection, and VaR/drawdown analytics
  • Gemini AI layer for structured stock reports, portfolio memos, and a chatbot with evidence-linked outputs and provenance tracking
  • No-trading compliance architecture — broker adapter has zero execution methods, guardrails sanitize AI output, 56+ pytest cases verify integrity
  • Next.js 15 dashboard with custom SVG financial charts, server-side data fetching, and Docker Compose across 50+ API endpoints
DigiMimic app icon

DigiMimic

Swift SwiftUI AVFoundation Metal Core Image Core ML Vision SQLite
  • SwiftUI camera UX bridged with AVFoundation so hardware capture logic stays out of UI views and preview state stays consistent
  • Metal/Core Image live preview through MTKView with 10-bit rendering and triple buffering for high-frequency frames
  • Still capture for HEIF, RAW/DNG, and RAW+processed output with portrait sidecars, PhotoKit integration, and deferred finalization jobs
  • Computational portrait and depth workflows using native depth, segmentation masks, and custom depth-of-field rendering with gallery refocus support
  • On-device gallery intelligence with MobileCLIP culling, duplicate clustering, color-match recipes, and SQLite sidecar persistence
The principles behind mobile craft, applied AI, and privacy-first architecture

How I
Build

Three pillars: polished mobile products, applied AI, and privacy-first architecture.

Mobile Product Engineering Shipped

Build polished mobile products

From interface systems to widgets, sync, exports, and device-native integrations.

React Native + Expo SwiftUI WidgetKit iCloud sync
Delivery focus End-to-end UX from screen flows to platform hooks.
Operating style Production-minded mobile systems that stay reliable beyond the demo.

Seen in: HabitForge and TrainTogether

AI, ML & Computer Vision Applied

Apply AI where it helps

Cloud models, on-device inference, and classical vision pipelines chosen by product constraint.

Gemini llama.cpp CoreML RoomPlan + ARKit
Product lens Latency, privacy, and workflow value decide the model strategy.
Proof On-device coaching in TrainTogether and spatial CV in Seekr.

Seen in: TrainTogether and Seekr

Privacy-First Systems By Design

Design around trust and offline utility

Local-first storage and narrow cloud usage so the product remains useful when privacy and connectivity matter.

Local-first storage Peer-to-peer sync Private iCloud container Offline-capable flows
Boundary setting Keep sensitive data close to the device whenever the product allows it.
Reliability Build workflows that still work under latency, weak signal, or strict privacy needs.

Seen in: Seekr and TrainTogether

Website case study

This portfolio is also part of the proof: a static product with optimistic UI, local fallbacks, and Supabase-backed interactions that stays polished when infrastructure is imperfect.

Static hosting Optimistic UI Local fallback logic
Capability Map

Project proof by system layer

Showing every shipped project and the capabilities each one proves.

TrainTogether Pose analysis, coaching, HealthKit, SQLite
Mobile AI/CV Privacy Native Testing
Seekr RoomPlan, ARKit, LiDAR, CloudKit sync
Mobile AI/CV Privacy Native
HabitForge React Native, widgets, Gemini insights, habit systems
Mobile AI Privacy Native
DigiMimic AVFoundation, Metal rendering, on-device gallery AI, SQLite sidecars
Mobile AI/CV Privacy Native Testing
IBKR Analyzer Read-only broker data, risk analytics, Gemini reports
AI Guardrails Testing Full-stack
Street, landscape, and light — the visual side beyond product engineering

Lens &
Light

Full Gallery
Street scene with atmospheric lighting
Mountain landscape vista
Candid street photograph
Architectural composition
Natural landscape photograph
Open to roles, collaborations, and conversations about the work

Let's Talk.

ryan@ryanyang05.me