Projects · 01 – 05

Five projects,
start to finish.

Open any of them for what it does, how it works, and where to get it.

01 / 05Habits & budgeting

HabitForge

Habit tracking and personal budgeting in one app, with nothing leaving the phone.

01 / 03

Habits and money together

The two things people track daily, in one app. Every record is stored on the device — no account, no server.

02 / 03

Logging has to be fast

Home Screen widgets and plain-English expenses — type “coffee $5” and it files itself.

03 / 03

AI stays optional

Analytics run locally. Gemini suggestions only appear if you supply your own API key, and everything still works without it.

Platform: React Native · Expo
Native edge: Swift widgets
Intelligence: Gemini · opt-in

02 / 05Room scanning

Seekr

Scan a room once, then find the thing you put away six months ago.

01 / 03

Walk the room once

LiDAR and ARKit turn the walkthrough into a 3D map. RTAB-Map holds the pose together so long scans do not drift.

02 / 03

Tag what is in it

Room, then furniture, then container, then item — the nesting matches how things are actually stored.

03 / 03

Find it again

Search runs on the phone and AR guides you back to the shelf. Works with no signal.

Capture: RoomPlan · LiDAR
Guidance: ARKit
Sync: Local first · CloudKit

03 / 05Fitness coaching

TrainTogether

Counts your reps and checks your form from the phone camera, without uploading video.

01 / 03

Watch the body

MediaPipe runs on the phone. Camera frames never leave it — that was the constraint the whole design had to satisfy.

02 / 03

Count the rep honestly

Real cameras shake and lag. Per-frame smoothing, a rolling window, and body-angle checks stop a wobble from being counted as a rep.

03 / 03

One scoring path, 32+ exercises

Every detector plugs into the same FormConfig and RepFormTracker, so adding an exercise does not mean rewriting the movement logic.

Vision: MediaPipe
Coach: Qwen · llama.cpp
Continuity: HealthKit · GRDB

04 / 05Camera

DigiMimic

A manual camera for iPhone, with the colour grade applied live instead of afterwards.

01 / 03

Real controls

Manual exposure, focus, and format through AVFoundation, plus zebra, waveform, false colour, histogram, and focus peaking.

02 / 03

See the grade before you shoot

Metal and Core Image apply the LUT in the live preview at 10-bit, with triple buffering and a degrade path when frames get expensive.

03 / 03

Keep the negative

HEIF, RAW/DNG, and graded video all come out the other side. Culling and colour matching run on-device; no photo is uploaded.

Capture: AVFoundation
Render: Metal · Core Image
Output: HEIF · RAW · DNG

05 / 05Finance dashboard

IBKR Portfolio Analyzer

A dashboard for my Interactive Brokers accounts that is architecturally incapable of placing a trade.

01 / 03

One view, several accounts

Holdings across accounts and both currencies resolve into a single position list with live P&L.

02 / 03

Say what the risk is

Allocation, VaR, drawdown, and concentration. Scoring is deterministic, and returns “Data insufficient” rather than guessing when inputs are missing.

03 / 03

The order path does not exist

The broker adapter has no order-placement methods, AI output is stripped of execution language, and 56+ pytest cases check it stays that way.

Engine: Python · FastAPI
Interface: Next.js
Guardrail: No execution methods

41 tools, and the project I used each one on. Click any of them.

Continue to Photography →

Photography · 144 frames

Photography

144 photographs in six collections — Toronto, the lakeshore, China, winter, Niagara, and the seasons turning. A camera habit that got out of hand.

See all 144 photographs →

Arcade · five games you can play

The arcade

Five games, each built around an algorithm I wanted to implement properly rather than approximate: expectimax, alpha-beta negamax, Hamiltonian cycles, constraint propagation, dancing links. Every one is honest about whether its AI can actually promise you a win.

Play them →

Games2048 · Connect Four · Snake · Minesweeper · Sudoku
SolversYou can watch them search
GuaranteesTwo of five can promise a win
InputTouch or keyboard

CS STUDENT
INDEPENDENT ENGINEER

OAKVILLE
TORONTO

ABOUT

Ryan Yang wearing a dark cap and sunglasses against a dark studio background.

Design, code,and support.

ABOUT / 05’26

WHO AM I.

I study CS at Toronto and build apps the rest of the time. Four are on the App Store. I did every part of them.

01 / WHO’S BEHIND THE WORKOAKVILLE, ONTARIO

Ryan Yang focused on writing code at a workstation. — RYAN YANGiOS / AI / SPATIAL

THE PERSON

I would rather implement the real algorithm than approximate it.That is most of the reason these took as long as they did.

METHOD / 02 INDEPENDENT ENGINEER / FOUR APPS THROUGH REVIEW

The demo is easy.
Week two is not.

METHOD

A pose tracker looks fine until the camera shakes. A scan looks fine until you reopen it a month later. Most of my time goes into the second case: smoothing, validation windows, restore paths, and what the screen says when there is nothing to show.

PRACTICE

I write the tests that would embarrass me if they failed — 56 of them just to prove one dashboard cannot place a trade.

Focus: iOS · Computer vision · Spatial
Studying: Computer Science, U of T
Shipped: 4 apps · 1 public repo
Based: Oakville / Toronto

CONTACT / 06 TORONTO · UTC-4 / EDT AVAILABLE FOR WORK

One email address, no form required

Tell me what you’re building.

A role, a product, or a problem you cannot get past — any of those is a good first email. A couple of sentences is plenty; I will ask about the rest.

Primary channel

ryan@ryanyang05.me

Optional

Prefer a form? These four fields do the same thing.

Four fields · opens your email app

Your name

Enter your name.

Your email

Enter a valid email address.

What is this about?

Tell me what you’d like to discuss.

Anything else worth knowing

Add a little more detail.

Nothing is sent from this page — your email app opens with the message filled in.

Software Engineer · Oakville / Toronto, Canada

I ship iOS apps, and the systems underneath them.

Computer Science student at the University of Toronto, and an independent engineer with four apps on the App Store. My work runs from ARKit and MediaPipe on the phone to FastAPI services and pytest suites on the other end.

I write for the case where the network is gone, the camera is shaky, or the data is missing — the states that decide whether people keep an app after week one.

View selected work Download résumé Email

01 / Experience

Where I have worked

2025 - Present

Independent Software Engineer

Oakville, Ontario

Mobile, AI, computer vision, spatial computing, and financial analytics

Designed, built, tested, and released four Apple-platform apps — habits and budgeting, LiDAR room scanning, on-device fitness coaching, and a professional camera — through App Store review each time.
Built a read-only investment dashboard end to end: IBKR Gateway integration, risk analytics, a deterministic scoring engine, and 56+ pytest cases that verify it cannot place an order.
Worked in Swift, SwiftUI, React Native, TypeScript, Python, and C++, across Apple frameworks, local databases, REST APIs, and Docker services.

Nov. 2022 - June 2024

Equipment Team Leader

East Asian Student Association

Coordinated event equipment, setup, and operational support for student programming.

2019 - 2021

Community Tech Support & Math Tutor

Oakville Chinese Network

Provided practical technology support and mathematics tutoring in a community setting.

02 / Projects

Selected production work

Five products I designed, built, and shipped on my own. Four are on the App Store; the fifth is a public repository. Each entry links to a longer case study and to the thing itself.

IBKR
PORTFOLIO
ANALYZER

IBKR Portfolio Analyzer

Independent Software Engineer · 2026 - Present

A read-only portfolio intelligence dashboard for consolidated Interactive Brokers holdings, risk analysis, deterministic scoring, and evidence-linked research.

Consolidates multi-account holdings with USD/CAD conversion, live P&L, allocation, VaR, drawdown, and concentration metrics.
Uses a broker adapter with no order-placement methods, guarded AI output, and 56+ pytest cases for no-trading compliance.

FastAPI · Next.js · TypeScript · Python · Docker · Gemini API · ib-insync

Case study → GitHub ↗

TrainTogether

Independent Software Engineer · Nov. 2025 - Present

An on-device fitness coach with live pose tracking, repetition counting, form scoring, session history, health integration, and local language-model coaching.

Built a MediaPipe landmark pipeline with smoothing, rolling-window validation, and body-angle constraints for stable real-camera analysis.
Designed shared FormConfig and RepFormTracker abstractions for 32+ exercise detectors and deterministic mock-landmark tests.

SwiftUI · MediaPipe · llama.cpp · SQLite · CloudKit · HealthKit

Case study → App Store ↗

Seekr

Independent Software Engineer · Sept. 2025 - Present

An offline spatial organizer that turns LiDAR and ARKit scans into searchable 3D room maps, so you can find where you put something months later.

Wrote a Swift-to-Objective-C++/C++ bridge that streams per-frame ARKit odometry into RTAB-Map SLAM while keeping mapping and export callbacks SwiftUI-safe.
Split persistence across Core Data metadata, canonical OBJ/PLY assets, and RTAB-Map databases so saved scans reload cleanly across sessions.

ARKit · RTAB-Map · C++ · Objective-C++ · Core Data · MultipeerConnectivity

Case study → App Store ↗

HabitForge

Independent Mobile Developer · June 2025 - Present

A cross-platform habit and personal-finance app covering streaks, budgets, reminders, widgets, analytics, and CSV export, with every record kept on the device.

Modeled 40+ domain entities across habits, finance, rewards, and notifications behind separate React Context providers and typed service layers.
Built a WidgetKit/App Groups batch sync through Swift and Objective-C native modules so habits, expenses, and budgets refresh in one payload.
Wired Gemini into recommendations and natural-language expense parsing on anonymized metadata, with caching so tracking still works with AI switched off.

Expo · React Native · TypeScript · Swift · Objective-C · WidgetKit · Gemini API

Case study → App Store ↗

DigiMimic

Independent iOS Engineer · June 2026 - Present

A landscape-first camera app with manual exposure and focus, monitoring tools photographers actually use, and colour grading applied live rather than after the fact.

Built the Metal/Core Image preview through MTKView with 10-bit rendering, triple buffering, and load-aware degradation when frames start costing too much.
Shipped capture paths for HEIF, RAW/DNG, and graded video, plus zebra, waveform, false colour, histogram, and focus peaking.
Kept gallery culling, subject segmentation, and colour matching on-device with Core ML and Vision; no media leaves the phone.

Swift · SwiftUI · AVFoundation · Metal · Core Image · Core ML · Vision

Case study → App Store ↗

03 / Education

Academic background

University of Toronto

Bachelor of Science, Computer Science Specialist

Faculty of Arts & Science · St. George Campus

Sept. 2024 – May 2028 (expected)

Cumulative GPA

3.73 / 4.0

Relevant coursework

Data Structures & Analysis, Software Design, Machine Learning, Systems Programming, Computer Organization, Probability & Statistics, Multivariable Calculus.

Honours & awards

Certificate of Distinction — Euclid Mathematics Contest Top 25% globally

2021 – 2024

Subject Award — Principles of Mathematics Highest grade in the course

2022

04 / Skills

What I work in

Languages Swift, Java, Python, TypeScript, JavaScript, C++, Objective-C++, SQL

Mobile & Apple SwiftUI, UIKit, React Native, Expo, ARKit, AVFoundation, Metal, Core Image, WidgetKit, Core Data, CloudKit, HealthKit

AI, CV & spatial MediaPipe, llama.cpp, Core ML, Vision, RTAB-Map SLAM, Gemini API, OpenCV

Backend & data FastAPI, Next.js, Node.js, PostgreSQL, SQLite, Docker, Supabase, REST APIs, ib-insync

Testing & tooling pytest, Playwright, XCTest, Git, Xcode, Instruments

Get in touch

I read every email. If you have a role, a product, or a problem in mobile, computer vision, or spatial computing, send the details and I will reply.

ryan@ryanyang05.me GitHub: RyanYang053 LinkedIn: ryan-yang

Four apps on the App Store. I built all of them myself.

HabitForge

Seekr

TrainTogether

DigiMimic

IBKR Analyzer

HabitForge

Habits and money together

Logging has to be fast

AI stays optional

Seekr

Walk the room once

Tag what is in it

Find it again

TrainTogether

Watch the body

Count the rep honestly

One scoring path, 32+ exercises

DigiMimic

Real controls

See the grade before you shoot

Keep the negative

IBKR Portfolio Analyzer

One view, several accounts

Say what the risk is

The order path does not exist

Photography

Photography

The arcade

Design, code,and support.

The demo is easy.Week two is not.

Tell me what you’re building.

Prefer a form? These four fields do the same thing.

I ship iOS apps, and the systems underneath them.

Where I have worked

Independent Software Engineer

Equipment Team Leader

Community Tech Support & Math Tutor

Selected production work

IBKR Portfolio Analyzer

TrainTogether

Seekr

HabitForge

DigiMimic

Academic background

University of Toronto

Cumulative GPA

Relevant coursework

Honours & awards

What I work in

Get in touch

The demo is easy.
Week two is not.