AI product advisory, engineering and assurance
AI that holds up in production.
We help organisations in Vietnam and Southeast Asia decide where AI pays off, build it on their own data, and keep it measurable, secure and accountable after launch.
Marc Nguyen, Founder and Principal10 years of product leadership. Ships AI to production.
How we work
From pilot to production, with controls at every step.
-
01 DiscoverDefine the business problem, the data available and the success metric in numbers.
-
02 ValidateRun a small, time-boxed test on your own data before any build commitment.
-
03 Build and measureDeliver in increments, each checked against an evaluation set drawn from real tasks.
-
04 Hand overCode, data, documentation and the evaluation suite stay in your accounts.
- Reference implementations
- 9
- AI systems designed, built and documented, each with a case study.
- Technical briefings
- 18
- Published analyses on evaluation, governance and retrieval.
- Services
- 8
- Each one backed by a system already running in production.
- Years in product
- 10
- Product leadership in EdTech and Fintech before AI.
Case studies
Reference implementations in production
Enterprise knowledge assistant
In productionSemantic search across 42,000+ documents in isolated workspaces, available in Claude on desktop, web and mobile, with cited sources.
- 42,000+ documents
- Workspace isolation
- Cited answers
Priority inbox triage
In productionAn agent that ranks email threads P0 to P2, verifies against project tools, and surfaces only items that require action.
- ~90 to ~8 min a day
- 568 to 7 priority items
- Human approval
MCP connector hub
In productionOne governed connector layer giving Claude access to Google Workspace, Slack, GitHub, Jira and Linear across three isolated workspaces.
- 150+ tools
- 3 isolated tenants
- OAuth and passkeys
Services
Advise. Build. Assure.
Advise
Decide where AI creates value and how success will be measured.
Build
Production AI systems on your data, delivered with the code and the evaluation suite.
- [03]
Custom AI Tools
Knowledge assistants and workflow agents on your documents and systems, delivered with source code. 4-8 weeks - [04]
AI Workspace Integrations
Connect Claude or ChatGPT to Google Workspace, Slack, GitHub and Jira through MCP, with tenant isolation. 2-5 weeks - [05]
Computer Vision
Detection and counting models trained on your images, running in a browser or on mobile. 2 weeks feasibility, 4-8 weeks to production
Assure
Measure quality, find security and privacy exposure, and verify usability and accessibility.
- [06]
AI Quality and Evaluation
Measure error rates in production and block regressions before release. 2-4 weeks - [07]
AI System Audit
Identify personal-data leakage, prompt-injection exposure and model-retirement risk in AI code. 1-3 weeks - [08]
UX and Accessibility Audit
Assess the product against Apple HIG and WCAG 2.2 with measured findings and remediation. 1-2 weeks
Works with the platforms your teams already run
- Claude
- ChatGPT
- Gmail
- Google Drive
- Slack
- GitHub
- Jira
- Cloudflare
- Swift
Insights
Insights and technical briefings
Start here
Have a problem AI might solve?
Describe the outcome you need. Within two working days you receive a written view on whether AI is the right tool, the risks to manage, and a proposed first step.