HVS Hanh Vuong Sy
Software architect · 2002

Hanh
Vuong Sy

I build the systems that keep ambitious products alive — ERP platforms, automation engines, data infrastructure, blockchain marketplaces, and AI tooling. Backend-first, end-to-end.

Backend systems AI agents Platform architecture
Explore

Scope of work.

4.5k GitHub commits Across 135 repositories
155 Pull requests Shipped and reviewed
7 Organizations Enterprise, Web3, AI, tools
135 Repositories Backend, AI, infra, clients

Domains of work.

D / 01

ERP & Enterprise SaaS

Enterprise platforms where a wrong field has real cost. I have shipped multi-tenant ERP backends for construction and supply-chain clients across quotation engines, bill-of-quantity calculation, procurement workflows, approval chains, and role-based access.

That means schema design that survives growth, workflow state machines with audit logging and soft-delete safety, custom formula evaluation, third-party CRM and accounting integrations, and the API contracts that let frontend teams move fast without breaking operational integrity.

Quotation & BOQProcurementWorkflowIAMAuditCRM / accounting sync
D / 02

Automation & AI Systems

Agentic workflows that do work while staying observable and controllable. I have designed RAG knowledge centres, code-generation agents with layered guardrails, and event-driven pipelines that carry signals from ingestion to action.

Execution is resilient by default: state machines for retries, fail-closed validation, semantic search over embedded documents, natural-language-to-code flows for formula repair, and clear surfaces where operators can intervene instead of guessing what the agent decided.

AgentsRAGCode generationSignal pipelinesGuardrailsZERO-SQL BI
D / 03

Data & Market Infrastructure

Market and analytics environments shaped for continuous inputs, interpretable signals, and action without losing context. I have built multi-timeframe data ingestion, order lifecycle tracking, journaling, risk surfaces, and real-time decision dashboards.

The architecture spans message queues and caching for throughput, vector storage for semantic retrieval, deterministic pipeline stages, and statistical / machine-learning exploration including RFM segmentation, clustering, and association-rule mining for customer intelligence.

Market dataOrder lifecycleAnalyticsReal-timeRiskCustomer mining
D / 04

Blockchain & Digital Assets

Product surfaces for marketplaces and blockchain ecosystems, designed around trust, lifecycle, liquidity, and the systems behind every transaction. I have worked on wallet flows, NFT marketplaces, order matching, royalty distribution, cashback, and payment orchestration.

Backend concerns include custody patterns, event indexing, transaction reliability, and the accounting surfaces that let users and operators trace every asset movement.

NFT marketplaceWallet & custodyTradingPaymentsRoyalties

Working stack.

S / 01

Languages

TypeScript · Python · Go · Java · C# · C / C++ · SQL · Swift · JavaScript (ES2022+) · Bash · Java ME (J2ME) · Assembly (NASM)

S / 02

Backend & API

NestJS (v10–11) · FastAPI · Gin · Express (v4–5) · Spring Boot · Next.js Custom Server · Prisma · Passport (JWT / OAuth / OIDC) · Azure MSAL · Stripe · Bull / BullMQ · Socket.io · SSE · gRPC · TCP microservices

S / 03

Data & Infra

PostgreSQL · Redis · RabbitMQ · Kafka · MongoDB · SQLite · MinIO · Qdrant · Docker / Compose · Nx Monorepo · Bull queues · Playwright · Cron

S / 04

AI & Search

Hermes Agent (custom code-gen) · LangGraph · LangChain · RAG pipeline · Qdrant (semantic search) · MCP Server · Langfuse · RAGAS · GEPA (evaluation) · Crawl4AI · faster-whisper

S / 05

Frontend

Next.js (v14–16) · React (v17–19) · Vite · Zustand · TanStack Query · Tailwind (v3–4) · Radix UI · Framer Motion · ReactFlow · react-signature-canvas · CRA · UMD

S / 06

Blockchain & Automation

Immutable X · Web3.js · TradingView CDP · MITM Proxy · OpenCV · OCR (Tesseract) · Selenium · adbutils · Chrome MV3 Extension · Telegram Bot API · Telethon · Kafka pipelines

Practice, not credentials.

System boundaries & service contracts

Nx monorepos split into API, web, engine, and IAM packages; TCP microservice communication; API gateways that own login, refresh, and JWT guard flow; centralized Prisma schemas shared across services. Worked across NestJS, FastAPI, Gin, Express, and Spring Boot backends.

Agent orchestration & knowledge workflows

Code-generation agents that write and repair source against a rule set, guarded by six fail-closed layers; RAG pipelines over vector stores with semantic retrieval; custom MCP servers; an AI proxy that routes across 40+ providers and reclaims 20–40% of tokens via real-time routing.

Data durability & real-time pipelines

Event streams over Kafka and RabbitMQ; Redis caching; Bull/BullMQ queues for campaign and background work; deterministic multi-stage pipelines with ordered steps; vector and object storage (Qdrant, MinIO) alongside PostgreSQL and MongoDB persistence models.

Product platforms & end-to-end ownership

Quote-and-BOM tools that reconcile formulas and catch unit and parameter errors before they ship; ZERO-SQL BI surfaces that query structured data without hand-written SQL; workflow builders and dashboards on React + Vite, delivered as Docker images with CI/CD.

Identity, access, and operational trust

Passport JWT, Google OAuth, and Entra ID with role-based guards; optional-JWT and auth interceptor patterns at the gateway; audit trails and fail-closed validation around anything that mutates business state.

Beyond product work.

Open source

DBX, OpenShip, Gantry and Hermes Agent.

8 merged pull requests to t8y2/dbx, a database client: streaming SQL Server ZIP imports, split-ZIP SQL import/export, relational-constraint bypass for Postgres and SQL Server, and ANSI SQL export fixes. Also submitted PRs to oblien/openship, getgantry/gantry, and NousResearch’s Hermes Agent, a self-improving agent system built around learning loops and experience-derived skills.

Applied learning

26 AI engineering labs.

Hands-on work across alignment, LLM fine-tuning, memory systems, RAG evaluation, deployment, guardrails, multi-agent MCP, and data foundations.

Let’s make the complex feel inevitable.