Projects

Key Features

• ▸Go + Gin API: JWT auth, RBAC middleware, Bloom Filter O(1) blacklist lookup
• ▸Multi-repo microservices: vshield-api, vshield-core, vshield-web, vshield-ios, vshield-android
• ▸iOS: Swift + CallKit real-time call screening & blocking
• ▸Android: Kotlin + Jetpack Compose + CallScreening API
• ▸Community reporting pipeline: Pending → Verified → Rejected with anti-abuse rules
• ▸Phone/bank account verification with community trust score aggregation
• ▸GitHub Actions CI/CD per repo: test + lint gate before merge
• ▸PostgreSQL 16 with JSONB + Bloom Filter index; Redis for rate limiting & session

Key Features

• ▸Django 5 + DRF backend with modular app structure
• ▸Next.js frontend with SSR/SSG (caheo-frontend repo)
• ▸GitHub Actions CI/CD pipeline (caheo-cicd repo): test → lint → deploy gates
• ▸Authored full API specification, data dictionary, and test case documentation
• ▸Multi-repo Document-First workflow: docs updated before code
• ▸Scrum workflow: Sprint Planning, Daily Standup, Sprint Review, Retrospective

Key Features

• ▸50+ REST endpoints: CapTable, FundingRounds, Shareholders, Financials, KPIs, Scenarios
• ▸JWT authentication: register / login / refresh / profile with RBAC authorization
• ▸Async CSV/Excel import (pandas + openpyxl): column validation, type checking, duplicate detection
• ▸Celery background workers processing 10,000+ rows; polling status endpoint for frontend
• ▸Swagger UI at /docs/ + Postman collection
• ▸Docker build optimization: 70–80s → 3–5s (95% reduction) via multi-stage layer caching

Key Features

• ▸Face attendance: ArcFace R50 512-dim cosine similarity (InsightFace buffalo_l model)
• ▸12-domain PostgreSQL schema: core, attendance, finance, health, face, auth, audit, billing…
• ▸PostgreSQL: Views, Triggers, Stored Procedures for billing & attendance automation
• ▸AI health risk scoring: Logistic Regression & Random Forest (AUC, F1, Brier evaluated)
• ▸FastAPI microservice: isolated face recognition via Docker internal network, zero port exposure
• ▸RBAC middleware (requireRole), rate limiting, CORS whitelist

Key Features

• ▸.NET 8 Clean Architecture: Api → Application → Domain → Infrastructure (4 strict layers)
• ▸EF Core migrations + auto-seed (roles, categories, sample data on first run)
• ▸JWT auth + RBAC: CUSTOMER / ADMIN / RESTAURANT_OWNER
• ▸Smart food recommendations: tag-based similarity + "frequently bought together"
• ▸Revenue dashboard with daily/monthly order statistics
• ▸React + Vite frontend with TypeScript

Key Features

• ▸3-site SQL Server horizontal fragmentation by phone prefix (MB: 096, MT: 091, MN: 090)
• ▸Linked Server cross-site queries: MB → MT → MN
• ▸Site MN = OCS: TaiKhoan (all-network) + TheCao management
• ▸Trigger tr_BaoVeGoiCuoc: validates subscriber existence across all sites before GoiCuoc deletion
• ▸Global revenue report: Promise.all([MB, MT, MN]) parallel queries with COUNT pushdown
• ▸Docker Compose: 5 containers (3 SQL Server + backend + nginx), auto db-init, 15× retry

Key Features

• ▸PDF upload → pdfjs-dist text extraction → 800-char overlapping chunks → Zustand localStorage
• ▸RAG Chat: keyword scoring → top-5 chunks → Gemini 1.5 Flash prompt injection → source citations
• ▸Auto Flashcard Generation: Gemini generates structured JSON Q&A pairs
• ▸SM-2 Spaced Repetition: 6-level rating (0–5), interval multiplied by easeFactor
• ▸MCQ Quiz: Gemini generates 4-option questions with explanations + live scoring
• ▸Gemini API key management with connection test; zero backend required

Key Features

• ▸Bilingual EN/VI via React Context (no external i18n library) — 300+ strings
• ▸localStorage language persistence across sessions
• ▸Next.js dynamic() lazy imports for all below-fold sections
• ▸Firebase Firestore contact form (lazy-loaded SDK on submit)
• ▸next/image priority + AVIF/WebP auto-conversion for avatar