Elvin Ong · Systems Portfolio

Senior Manager, E-Commerce · Kino Biotech (Kinohimitsu), Singapore

o.jincheng@hotmail.com

+65 8698 7701

Two internal tools I conceived and co-built at my current employer, taken from a business problem to something the team uses every day. Both were built in-house with AI-assisted development, without an external agency.

Everything on this page is an illustrative reconstruction. Interfaces, imagery and figures are examples, not the employer's production interfaces, source code or company data. Both systems are company-owned.

Case 1 · AI Content Engine

Built the operating system behind the content

Creative testing on TikTok Shop, Shopee, Lazada, Meta and Google was limited by how fast an agency could produce assets. The engine turns one product input into a controlled stream of videos, images and ads that the content and media teams can deploy directly.

1,000+

Deployment-ready assets a month

0→1

Conceived, built and rolled out in-house

Lower

Unit cost vs agency production

My role Product concept · Workflow design · AI-assisted development · Internal rollout · Performance feedback loop

Not just generation: a controlled workflow

1InputProduct assets, campaign goal, format and audience

2GenerateMultiple hooks, scenes, formats and variations

3ControlProduct scale, brand and voice guardrails applied

4ReviewHuman QA for visual detail, captions and usability

5DeployApproved assets released to content and media teams

Deployment problems we had to solve

The problem

What we built

Product looked too large or unrealistic in frame

Automated product-scale and framing control

Voice delivery or pronunciation sounded unnatural

Pronunciation and tone review built into the workflow

Brand visuals inconsistent across outputs

Brand rules and asset control enforced automatically

Product details unstable or distorted

AI-plus-human consistency checks before release

Captions or lip-sync felt unnatural

Dedicated QA pass for sync accuracy and captions

Output range from one product input

Official brand videoPolished, campaign-ready cuts
UGC-style contentAuthentic, relatable, trust-building
Product imageStudio-quality visuals at scale
Social adScroll-stopping, performance-tuned
Short-form videoMore formats, more reach

TikTok live commerce and AI short-drama content

Sample content formats. Product shown is my own side-project brand, used here as the illustrative input.

Business value

Create moreHigher testing volume without a matching increase in production workload.

Learn fasterCreative-level signals connected to media and commerce performance.

Improve controlBrand rules and human review cut unusable outputs before deployment.

Scale the teamA hands-on build turned into a repeatable workflow the wider team runs.

What this shows: the difference between a one-off AI experiment and a production system. Guardrails, QA and a feedback loop that a team can trust and run without daily supervision, expanding creative testing capacity well beyond what agency production could support, without a matching increase in headcount or spend.

Case 2 · Commerce Intelligence

The dashboard behind the decisions

Eight channels used to mean eight manual reports and three people compiling them. The dashboard pulls daily sales, traffic, ad and SKU data into one ranked view, then an LLM API call writes a short daily summary of what is working, what is not, and what to do next.

Channels consolidated

Daily

Automated sync

Report owner (was 3)

Real-time

Channel ranking

My role Requirements · Data model and channel mapping · AI-assisted development · Management rollout · Daily use as the reporting owner

Revenue share by channel (illustrative)

TikTok Shop

31%

Shopee

25%

Lazada

19%

Shopify (DTC)

14%

iShopChangi, KrisShop, RedMart, S-Mart

11%

Top performing lines, 7-day (illustrative)

Range	Share	WoW
Immunity & Wellness Range	22%	+9%
Collagen & Beauty Range	17%	+14%
Probiotic & Gut Health Range	13%	−3%
Multivitamin Range	10%	+2%

AI summary and recommendation

Auto-generated daily

"Collagen & Beauty range GMV is up 14% WoW, driven by TikTok Shop live sessions, while the Probiotic range dipped 3% as CAC rose on Meta. Recommend shifting ~10% of Meta prospecting budget to TikTok Shop and refreshing the Probiotic range's top creative, which is now 3 weeks old."

Example output, generated via an LLM API call over daily traffic, sales, ad and SKU data. Figures are illustrative, not real performance data.

What this shows: commercial judgment, product thinking, hands-on AI execution and team adoption in one person. It is the same instinct behind the results on my resume: find the constraint, design the system, ship it, then scale it through the team.