Product is
Response
TO A WELL-DEFINED PROBLEM.
AUTHOR: ANDREW DO
DOMAIN: ENTERPRISE · EDTECH · AI
BASE: SEOUL, KR (OPEN TO RELOC)
STATUS: OPEN TO SENIOR PM ROLES
> Projected 80% cost reduction on a 1.5M-MAU e-commerce overhaul.
> Built two production SaaS products from scratch. $0 budget.
terminal / 02_System_Specifications
[READ_ONLY]Professional Experience
- >> Own APMEA North's digital product roadmap across Glo & Velo — primary liaison between Japan/Korea local markets and BAT global engineering.
- >> Defined frontend architecture spec for Japan's Adobe Commerce migration; designed cost model targeting -80% op. costs YoY.
- >> Built bat-ly end-to-end — a @bat.com-gated URL shortener now serving BAT Korea's full marketing team, eliminating data governance risk and preserving 100% UTM attribution.
- > Architected and pitched Ministry of Education AI platform proposal — led cross-functional team that secured ₩90B (~$65M) national research contract.
- > Grew K-12 AI coding platform from 0 to 50,000 monthly active users through government distribution channels and curriculum partnerships.
- > Led QA for TikTok Korea's content safety pipeline — designing and executing test protocols across multiple content categories at scale.
- > Identified and escalated UX inconsistencies across localised mobile experiences, contributing to product fix cycles aligned with global platform standards.
- > Coordinated multi-vendor scheduling and delivery milestones for Samsung Electronics' £multi-million UK enterprise digital transformation — engaging procurement, systems integration, and compliance stakeholders across Korea and the UK.
Technical Stack
// Enterprise
// Product (Personal)
Management Skills
- REQ DEFINITION
- STAKEHOLDER MGMT
- X-FUNCTIONAL COLLAB
- UAT MANAGEMENT
- DATA ANALYTICS
- AGILE / SCRUM
Certificates
Education_Log
Glo/Velo Japan
E-Commerce
As lead PM, I drove a 1.5M MAU platform migration off a fragmented legacy stack onto BAT's global Adobe Commerce standard.
Problem
Japan's Glo/Velo platform ran on a fragmented, market-specific CMS stack incompatible with BAT's global engineering standards. Cross a 6-person team spanning Japan marketing, BAT global engineering, and Fujitsu systems integration — siloed components and manual promotion deployments were inflating operational costs by an estimated 80% above global baseline. The migration had to happen without disrupting active campaigns.
Solution & Impact
Led frontend architecture redesign — spec'd 40+ reusable UI components, authored the migration PRD, and coordinated phased UAT with local market stakeholders.
Platform successfully migrated. CI/CD adoption eliminated manual promotion downtime. Global component library now enables cross-market UI reuse that previously required separate development cycles per market.
Related Deployment
>> BAT_KOREA D2C LAUNCH
Led architecture for BAT Korea's multi-brand D2C market entry: integrated Adobe Commerce, Salesforce CRM, and logistics systems to support complex membership & compliance.
Public SaaS
Scaling
AI-integrated K-12 digital textbook platform. Ministry of Education.
Problem
South Korea's national curriculum digitisation opened a competitive government bid requiring an AI-powered K-12 textbook platform that met strict safety and pedagogical compliance standards. No market-ready solution existed. Elice's window to win was narrow — and the contract would define the country's primary school digital infrastructure for years.
Solution & Impact
Led the Ministry of Education pitch strategy, compliance documentation, and cross-functional product design — positioning Elice as the only platform capable of meeting both AI capability and national safety requirements.
Won a ₩90B 3-year national research contract. Scaled the K-12 platform from early users to 50,000 monthly active users, distributed through government curriculum channels and institutional partnerships.
bat-ly
Identified a compliance gap. Wrote the PRD. Built and shipped to production — solo, in 6 weeks.
Problem
Long UTM campaign URLs caused link-rot and readability issues across BAT Korea's marketing channels. External shorteners (bit.ly) violated BAT's corporate data governance policy — any click data would leave the company's perimeter, putting campaign analytics at risk.
Solution & Impact
Owned the full product lifecycle — PRD authoring, system architecture design, and production deployment — without a single external engineer. Gated by @bat.com corporate email via Supabase Auth.
100% UTM parameter preservation via Vercel Edge Middleware. Async click analytics (referrer, device, country) captured without blocking redirects. CSV bulk upload handles 100 links/batch. Now serving BAT Korea's entire marketing team.
Technical Deep Dive
>> EDGE_RUNTIME
Vercel Edge Runtime handles DB lookup, fire-and-forget click logging (Promise.all), and redirect in a single ultra-low latency step.
→ Zero latency overhead. Analytics captured without slowing users down.
>> ROW_LEVEL_SECURITY
Access control enforced at the PostgreSQL policy layer via Supabase Auth + RLS — not just in application code.
→ Data governance policy satisfied. IT team involvement: zero.
>> PRODUCTION_DASHBOARD
InvestMate
A fully automated AI investment analyst. Solo-built. Delivered to Telegram every morning. $0 infrastructure cost.
Problem
Real-time, personalised market intelligence doesn't exist for retail investors — at least not one that integrates your actual portfolio, runs daily analysis on macro conditions and earnings, and delivers a concrete BUY/HOLD/SELL plan with zero manual input. I built it to prove that the same product instinct I apply at work translates directly into shipped AI infrastructure.
Solution & Impact
Designed and shipped a serverless AI pipeline — GitHub Actions cron triggers Gemini 2.5 Pro analysis of macro conditions, earnings data, and sector news. Results are pushed as a structured BUY/HOLD/SELL plan directly to Telegram every morning.
Fully automated: zero-touch daily execution. Insights delivered in under 30 seconds via Telegram — no app required. Running at $0/month on Vercel + Supabase free tier infrastructure. Same product design process I apply on-the-job, applied to a live personal AI product.
Tech Stack
>> AI_ENGINE
Gemini 2.5 Pro for macro analysis · Gemini 2.5 Flash for real-time news translation
>> AUTOMATION
GitHub Actions cron → Vercel Serverless → Telegram Bot API. Fully hands-off daily pipeline.
>> SECURITY
Google OAuth + Supabase RLS. Admin-controlled whitelist. One-time pairing tokens for Telegram (10-min TTL).
>> LIVE_INTERFACE
Next
Problem?
If you've got a product problem that's costing money, losing users, or blocking scale — let's talk.
INITIATE_CONTACTSEOUL, KOREA
KST (UTC+9)