Software Engineering student in Cambodia building a private startup portfolio across market access, education, cybersecurity, SME operations, trading tools, and student opportunity infrastructure, alongside public AI-native developer tooling.
I am early in my career, but I am deliberately training like a product engineer: ship working systems, document tradeoffs, protect user trust, and use AI agents as engineering leverage rather than as a substitute for judgment.
- Building seven startup-track products under a Cambodia-first product lab
- Turning local market problems into small, testable software systems
- Practicing AI-native engineering: specs, agent reviews, test loops, and clear implementation logs
- Publishing reusable open-source workflow tools when they can help other student and early-stage teams
- Keeping future-startup source code private while publishing honest public proof of work
- Portfolio case studies: kavatana.me documents the product surfaces, screenshots, tradeoffs, and build direction behind my work.
- Studio OS demo: kavatana-studio-os.vercel.app is the neutral public mode of a private AI-native command center. The public build keeps state local and disables live connectors; the private operator adds bounded workspace context and a read-only GitHub organization queue. The human-agent control-plane case study documents its authority, context, privacy, and verification boundaries.
- Portfolio concierge: the site-wide guide answers and routes from committed public evidence in the visitor's browser. It cites approved sources, keeps session context in the current tab, makes no model or private-tool call, and follows the city-wide concierge contract.
- CHNAI LAB profile: github.com/chnai-lab presents the student-run product studio behind the broader portfolio.
- Collaboration system: CHNAI LAB repositories use canonical agent guides, contribution and security boundaries, team ownership, and CI so human and AI collaborators enter through the same reviewable workflow.
ai-native-team-starter
is a vendor-neutral GitHub template for teams using coding agents without giving
up human ownership, security boundaries, or verification discipline. The public
v1.0.0 release
includes agent instructions, issue and pull request forms, onboarding and
security guidance, a zero-dependency readiness scanner, regression tests, and
pinned GitHub Actions verification.
Read the open-source engineering case study for the problem, design decisions, safety boundary, and verification evidence.
This profile is also a public Nuxt application, not only a narrative README. Its repository policy, claim boundaries, content schema, media references, concierge behavior, credential patterns, and pinned CI dependencies are checked by a zero-dependency Python verifier with regression tests.
npm ci
npm run verifyThe same command runs in GitHub Actions and produces the static portfolio build.
Most source repositories are private because these are intended startup assets, not throwaway class projects. Public details are documented at a product and architecture level.
| Product | Problem Space | Current Thesis |
|---|---|---|
| BayonHub | Student opportunity infrastructure | Help Cambodian tech students find internships, freelance work, jobs, and portfolio-building opportunities. |
| Svaeng Yul | Khmer tech education | Khmer video self-study for AI engineering, data science, cybersecurity, software engineering, and related fields. |
| Chomkar | Agricultural market access | Connect buyers and farmers before harvest demand is locked in, reducing hidden coordination power from middlemen. |
| Sat Digital | Community cybersecurity | Rentable cyber guard tools for Telegram communities, groups, and websites. |
| Vantrex | Trading software | Signal and indicator tooling across Forex, crypto, and model-assisted trading workflows. |
| PHSAROS | SME operations | Operating system for restaurants, coffee shops, marts, staff, inventory, and finance. |
| CHNAI LAB | Student-run product studio | A Cambodian student team building owned products with real responsibility, equity, and operating discipline. |
Read more:
- Startup portfolio map
- Studio OS human-agent control plane
- AI concierge and product-agent city
- Open-source engineering proof
- AI-native engineering workflow
- GitHub account cleanup record
- Repository agent contract
TypeScript, JavaScript, Vue, Nuxt, React, Next.js, Node.js, Python, GitHub GraphQL, Supabase, PostgreSQL, Cloudflare, Vercel, GitHub Actions, and AI-assisted development workflows.
- Product first: start from a real user, workflow, and failure mode.
- Evidence over hype: document what is shipped, what is simulated, and what still needs validation.
- Private by default for startups: protect source code, secrets, and product strategy until a project is ready for public release.
- AI-native, not AI-blind: use agents for speed, review, research, and implementation support, then verify outputs with tests, code review, and manual judgment.
- Bounded authority: give agents the minimum context and permissions needed, and keep consequential product, security, deployment, and financial actions under human control.
- Readable systems: prefer clean architecture notes, small commits, and decisions that a future teammate can understand.
- Portfolio: kavatana.me
- GitHub: github.com/kavatana
- LinkedIn: Ka Vatana