Another vendor couldn't deliver this OTT platform. We took it over and got it running — then they gave us the app too
NTDVideo's private OTT platform was stuck because the previous vendor couldn't deliver. Brightstream took it over, fixed it, and got it fully live and running normally. Then the client gave Brightstream the app too.
- Client
- NTDVideo OTT
- Industry
- Media / OTT
- Service
- Custom web & apps
- Led by
- Johnny Yang

What we delivered.
- Time to launch
- 18 core features delivered in 3 months
01I took on a project someone else couldn't deliver
NTDVideo (新唐人, Xin Tang Ren) is a media company producing high-quality original video. What they wanted was clear: their own private OTT platform — no more parking their audience and their data on someone else's public platform.
The problem: by the time it reached me, the project was already stuck.
The previous vendor couldn't finish it and couldn't deliver. A video platform involves members, payments, permissions and high-concurrency traffic. Getting stuck halfway is the hardest state to be in: you don't just have to keep building, you first have to figure out which parts of what was left behind don't work.
A lot of people won't take rescue projects. They're high-risk and thankless. I took it.

02What we did: get it working first, then get it right
I brought my team in to take over and rebuild the platform properly — one ASP.NET Core 8.0 multi-project solution, 5 projects tied together into a complete OTT system:
| Project | Role |
|---|---|
| NTDVideoTw.Web | Member front end (Razor Pages, 17 front-end pages) |
| NTDVideoTw.Staff | Admin back office (Telerik Kendo UI, 9+ modules) |
| NTDVideoTw.Api | API service (member / notification / video / subscription endpoints) |
| NTDVideoTw.FipSvc | Payment service (NewebPay integration, deployed separately) |
| NTDVideoTw.Shared | Shared models / Services / Utils |
The database is RavenDB (NoSQL). Video metadata, members, subscriptions and notifications are all document-oriented. That avoids the schema hell a traditional relational DB hits in a many-to-many scenario like "members × videos × watch progress × favourites × notifications". The cloud-native architecture is deployed on AWS EC2 and handles high concurrency.

03We solved a very practical Docker deployment pain point
The OTT back office is heavy to compile. NTDVideoTw.Staff alone (Telerik Kendo plus back-end logic) ran out of memory (OOM) compiling inside the container.
Our fix: a pre-build strategy — move compilation out of the main docker-compose.yml into docker-compose.build.yml, build the images on the host first, then deploy. docker-compose up pulls the ready-made ntdvideo-web:latest / ntdvideo-staff:latest, so deployment doesn't blow up memory.
This kind of deployment detail is what textbooks don't cover but will kill you. We turned it into an SOP, and future .NET clients at a similar scale can reuse it directly.

047 core features, 100% live
This project had a defined delivery list of 7 core features. All went live in the first half of 2026:
- New-video member notifications (on-site notification system, auto-push with thumbnails)
- Member notification center (real-time read/unread, with Watch Later and Favourites separated)
- Video release dates (shown on a front-end timeline)
- OAuth with 4 providers (Google / Facebook / LINE / Apple)
- Staff back office (video management / member management / subscriptions / reports)
- FipSvc payment service (NewebPay recurring subscription billing, deployed separately to reduce risk)
- API service (for a future app / WebView)
Overall completion is 93%. The remaining gaps are the "mobile app" and "fixed-IP payments" (planned, but the client didn't start them in that phase).

05The duplicate-charge incident — how we handled it
In 2026-03 the client reported that some members had been charged twice by NewebPay. What we did:
- Same day: pulled every duplicate transaction record (NewebPay Period Detail CSV)
- Next day: found the root cause (missing webhook idempotency + a client-side race condition on the client's side)
- Day three: fix + data cleanup (every wrongly charged member refunded in full)
- Day four: wrote it up as an internal SOP,
2026-03-01-訂閱重複扣款修復與資料清理.md
A payment bug is fatal for a subscription business. We didn't explain after the fact. We stopped the bleeding the same day, fixed it the next day, and returned the money on day three. The client didn't lose a single member.

06Then they gave us the app too
Once the rescue was done, the same thing happened as with every good project I've had: the client came back and gave Brightstream the next, more important project — WebView app development (iOS + Android, launched together), including a member referral code system + a discount code system, on an 8-week timeline.
Real delivery isn't "finishing a project". It's when the client, after seeing how you close things out, is willing to put something bigger in your hands.

07The real value of this project: from traffic thinking to asset thinking
The tech is only the surface. Underneath, the NTDVideo project is a shift in how a media business thinks.
On a public platform, you have traffic, but the audience isn't yours and the data isn't yours. The platform changes its policy and your audience can disappear. The point of a private platform is to turn "traffic" back into "assets" — viewers become your own members, behaviour becomes your own data, revenue becomes your own subscription cash flow.
This holds for anyone who makes a living from content: what you need isn't more traffic. It's users who actually belong to you.

08What Brightstream delivered
| 項目 | 內容 |
|---|---|
| Platform status | From "previous vendor couldn't deliver" → fully live, running normally |
| Tech stack | ASP.NET Core 8.0 multi-project (5 projects) / RavenDB / AWS EC2 |
| Staff back office | Telerik Kendo UI / 9+ back-office modules |
| Payments | NewebPay (recurring subscription billing, FipSvc deployed separately) |
| 7 core features | 100% complete |
| Overall completion | 93% |
| Notification system | On-site + real-time read status + Watch Later + Favourites separated |
| OAuth | Google / Facebook / LINE / Apple, four providers |
| Deployment | Docker pre-build (fixes OOM) + maintenance-mode deployment with no downtime |
| Follow-on | Client commissioned more: WebView app + referral code system + discount code system (8 weeks) |
09What NTDVideo says
"When the project was stuck, we went to Brightstream. They didn't just clean up the earlier problems — they made the whole platform solid. Later, when the subscription billing went wrong, they stopped the bleeding the same day, fixed it the next day and returned the money on day three. That quality of follow-through is why we handed them the app directly." — NTDVideo (drafted by Brightstream, approved for publication)
10From Johnny Yang
A lot of vendors won't take rescue projects, because you end up responsible for problems someone else left behind. But I'm willing to take them — because how a company finishes a project tells you more than how it starts one.
NTDVideo later gave us the app. Not because my pitch was good. Because they saw with their own eyes how we cleaned up a mess, and how we handled a payment incident. B2B trust isn't won in a pitch deck. It's earned in how you close things out.
When the project was stuck, we went to Brightstream. They didn't just clean up the earlier problems — they made the whole platform solid. So when it came to the app, we handed that straight to them.

