06/2026 – 09/2026, SMAC Leverage
- Role
- Engineering Manager
- Team
- 5-person core team, plus an 8-person vendor
- Website
- saora.ai
Business goal
A multichannel AI marketing platform for Vietnamese SMEs selling on social media, readied for its first commercial wave.
SAORA.AI is version 2 of SMAC’s AI Marketing product: a multi-tenant SaaS that helps Vietnamese shops selling on social media plan campaigns, produce content and look after customers. The core rule is that AI proposes and a person approves; nothing is posted automatically.
Main modules
- Campaigns: a five-step wizard
- Content studio: copy variants, AI images, 15-second video, brand-safety and advertising-law checks
- Unified inbox, CRM and pipeline
- Analytics
- AI governance: approvals, permissions, audit log, AI cost metering
- Billing and credits (MoMo, bank transfer)
- Super-admin portal: 17 screens, 93 states
My part
Took SAORA.AI to the go-live line: scope locked at 8 modules and 218 screen states, the vendor held to 5 release blockers, and an 18-month financial model put to leadership.
- Own the roadmap, the scope of each release and the go-live readiness criteria
- Run a five-person core team and the vendor on one-week cycles and a single board
- Manage the vendor, AI cost and cloud cost
- Define quality criteria and acceptance conditions before each release
- Report progress, risk and the financial model to leadership
Result
- 218approved screen states as the scope, across 8 modules
- 51 → 1work items moved onto one board; the legacy tracker was closed
- 10/2026planned go-live, after I handed the project over in 09/2026
Problems faced
The backlog was spread over a legacy tracker with two conflicting status fields, so nobody could read real progress. Most of the product was built by a vendor, so the risk sat in scope and acceptance criteria that were not yet sharp enough to hold the vendor to. Leadership also had no financial basis for the launch decision.
How it was solved
- Moved 51 work items from Odoo onto one Plane board, locking a one-to-one status mapping and 15 labels before the move.
- Locked the scope at 8 modules and 218 approved screen states (125 tenant-side, 93 admin-side).
- Set five release blockers and UAT with real tenants, so the vendor could not go live before the product was ready.
- Built the HTML prototype on Chakra UI v3 and the saora.ai website myself, with Lighthouse CI gates on accessibility, SEO and performance.
- Prepared the leadership pack: an 18-month financial model, a 90-capability comparison with competitors and the trial package unit economics.
- Advised leadership on the commercial team structure ahead of launch.
Project management process
- 1Kick-offKick-off decision, vendor and five phases locked
- 2Scope and design8 modules, 218 screen states, Chakra UI prototype
- 3One-week cyclesMonday planning, Friday demo and retro
- 4Gate reviewsGates 0–4, weekly reports to the steering committee
- 5UATUAT-1 on 05/09, UAT-2 on 02/10 with real tenants
- 6Go-livePlanned for 10/2026, once the 5 release blockers are met
- Hybrid: the foundation on a fixed plan, development in one-week Scrum cycles (Monday planning, Friday demo and retro, 15-minute daily).
- Gates 0 to 4 with weekly reports to the steering committee; weekly KPIs on progress by feature ID, open bugs and AI cost per tenant.
- An outside vendor delivered on a fixed-price contract in five phases, held to five release blockers: AI content labelling, tenant isolation, payments, AI cost per tenant and data-protection paperwork.
- Requirements clarified with the vendor in numbered rounds; a deliverable passed only when both sides confirmed it.
- UAT with at least five real tenants, plus a penetration test before go-live.
- A nine-document legal and compliance pack aligned with Law 91/2025/QH15.
Management stack
- Hybrid
- Scrum
- Plane
System architecture
- Users
- Tenant app
- Super-admin portal
- saora.ai website
- Services
- Campaigns
- Content studio
- Inbox + CRM
- Analytics
- AI governance
- Billing
- AI
- OpenRouter
- Gemini + Veo
- n8n workflows
- Safety gate
- Integrations
- TikTok
- YouTube
- Zalo OA
- MoMo
- Infrastructure and DevOps
- AWS
- Google Cloud
- 3 environments
- Feature flags
Tech stack
- AI
- OpenRouter
- Gemini
- Veo
- n8n
- Channels
- TikTok
- YouTube
- Zalo OA
- Payments
- MoMo
- Bank transfer
- Viettel S-Invoice
- Cloud
- AWS
- Google Cloud
- Design and prototype
- Chakra UI v3
- Figma
- Playwright
- saora.ai website
- Next.js 16
- React 19
- TypeScript
- Tailwind v4
- Docker
- GitHub Actions
- Lighthouse CI