Skip to main content
Charles Tsoi

AI · Machine Learning · App Development

Charles Tsoi

AI · ML · App development · Apple · Accenture · AAHK

Experience with AI agents, machine learning, and app development — happy to help where I can contribute

Introduction · Career story

A consistent thread.

Product → architecture → AI

Thanks for having me. Here's a short version of how I got here — on paper it can look like a few jumps, but there's a fairly consistent thread.

The thread is this: over fourteen years I've moved from digital product delivery, to systems architecture, to AI — usually sitting between business stakeholders and engineering teams, helping translate what one side needs into what the other side builds.

Mobile product ownership

2012 – 2018

Six years at Cathay Pacific.

Where product ownership fundamentals took root

12x

Mobile MAU growth (~10k → ~120k)

6 years

Lifecycle ownership

NotiFLY

Modernised notifications

I started at Cathay Pacific — six years as a Product Owner on their mobile app and related notification systems. During that time the app grew from roughly 10,000 to 120,000 monthly active users. That's where I learned backlog discipline, stakeholder negotiation, and how to integrate new frontends with older backends like PSS.

Six years at one company also taught me what it means to own a product across its lifecycle — not only launch, but maintain and evolve it.

  • Product ownership on Cathay Pacific Mobile App — MAU ~10,000 → ~120,000 (12x)
  • Defined XML/JSON structures integrating mobile frontend with legacy Passenger Service Systems (PSS)
  • Helped revamp NotiFLY — migrated legacy notifications toward a modern messaging architecture

Regulated real-time systems

2019 – 2022

Regulated systems. Careful architecture.

HKJC + Accenture

Solace

REST → event-based messaging (HKJC)

20%

Less integration rework (API contracts)

Zero downtime

Regulated real-time systems

From there I moved through HKJC and Accenture. At Jockey Club, I worked on a Solace migration — moving core infrastructure from traditional REST APIs toward event-based messaging for real-time, high-concurrency betting in a regulated environment.

At Accenture, I spent more time on the architecture side — API contracts and helping teams shift from feature factories toward value-driven delivery.

  • Solace migration at HKJC — traditional APIs toward event-based messaging for real-time odds
  • API contracts (Swagger) reducing integration rework by ~20% (Accenture)
  • Helped shift mobile squads from feature factory toward value-driven architecture

Edge & serverless delivery

2022 – 2024

A standard, not only a technology.

iWork serverless · Genmoji delivery

Edge

docserverless — compute on devices

Serverless

iWork backend migration

Genmoji

Cross-team API delivery

Then Apple — two years as Engineering Project Manager, including work on the docserverless edge project, which moved numerical computation from server-side to user devices. What I took from Apple wasn't only a technology — it was a bar for how carefully decisions were reviewed, dependencies mapped, and releases evidenced.

That kind of care is what I try to bring into regulated, high-stakes environments.

  • docserverless — numerical computation moved from server-side to edge user devices
  • Partnered on iWork backend serverless migration — architectural reviews & dependency mapping
  • Helped unblock API dependencies across CoreML, UI Engineering, and QA

AI in production

2024 – Present

AI in production. Scope as a skill.

Airport Authority Hong Kong · 2 years

~80%

Estimated less manual reporting time

10 weeks

Capacity planning AI delivery

Scope

Gatekeeping in practice

At Airport Authority I've been involved in product design and vendor delivery for LLM-based agents in production, vendor technical due diligence, and keeping scope honest. One example: an infrastructure monitoring agent (Power Automate + Dynatrace + ServiceNow) that reduced manual health reporting for senior management.

  • Product design and vendor delivery for an LLM monitoring agent — estimated 80% reduction in manual reporting time
  • Vendor technical due diligence on Microsoft ecosystem — API, security, scalability
  • Delivery support on a new Insurance System — legacy mainframe to cloud migration
  • Hands-on: LungLens live at lunglenshk.vercel.app — see Work page

Selected GitHub work →

Continuous learning

Education & momentum

Still learning.

CUHK MSc · Continuous learning

I'm finishing my MSc in IT Management at CUHK, steering coursework toward AI and data science so I have more hands-on technical grounding behind the product and architecture work I do day to day.

Enterprise delivery experience plus formal study in data science and system design is how I try to stay useful across product, architecture, and AI initiatives.

  • MSc IT Management, CUHK — focus: Data Science, AI, System Design (Expected Q1 2027)
  • Certifications: PMP, PSPO I, PSM I, Solace Certified Event-Driven Architecture Practitioner
  • Hands-on repos and production work keep skills current alongside formal study

What I can contribute

  • AI & ML: LungLens in production; AAHK LLM monitoring agent — product design and vendor delivery (estimated 80% less manual reporting time)
  • Product: Fourteen years across Apple, HKJC, Cathay, AAHK — Cathay mobile MAU ~10k → ~120k (12x) during my ownership
  • Architecture: Solace event-driven work at HKJC; docserverless edge compute at Apple
  • Building now: 北辰 (North Star) astrology app for iOS; writing on GenAI systems on Medium

Let's connect.

Open to collaborations, consulting, and full-time roles. If you have a problem in AI, ML, or app development, I’d be glad to talk.