Portfolio · GitHub · RAG API · Commerce demo
I build evidence-grounded AI applications and the software systems around them: retrieval pipelines, APIs, web and mobile products, and edge integrations. My work favors explicit contracts, security boundaries, reproducible tests, and honest failure states over unsupported claims.
Computer Science Engineer based in Cần Thơ, Việt Nam, open to AI Engineering, Mobile, and Software Engineering opportunities.
- AI Engineering — Python, PyTorch, FastAPI, NLP, hybrid retrieval, vector search, Retrieval-Augmented Generation, semantic claim verification, and recommendation systems.
- Backend & system design — Laravel, PHP, Node.js, NestJS, TypeScript, REST APIs, authentication, authorization, idempotency, rate limiting, and transactional data integrity.
- Frontend & mobile — React, Next.js, TypeScript, Vite, Flutter, Dart, Riverpod, accessible interfaces, and client-side state management.
- Data & infrastructure — MySQL, PostgreSQL, MongoDB, Redis, Qdrant, Docker, Linux, GitHub Actions, and automated browser/API testing.
- Edge AI & IoT — Raspberry Pi, ESP32, MediaPipe, OpenCV, MQTT, PlatformIO, Firebase, and embedded C++ integration.
- Blockchain integration — Solidity, Hardhat, ethers.js, IPFS, QR-based verification, and smart-contract-backed state.
Hybrid academic search and grounded question answering over research-paper metadata. The system combines lexical and dense retrieval, streams answers over SSE, validates citations, and maps factual claims back to retrieved evidence.
Python FastAPI Qdrant BM25 RRF Streamlit ·
API documentation
Status: research/demo system; semantic verification reduces grounding risk but does not guarantee factual correctness.
A multi-client commerce system covering a Flutter customer app, React storefront, administration portal, and server-authoritative APIs. The work includes Sanctum authentication, capability authorization, inventory controls, idempotent checkout, COD/SePay boundaries, and automated browser tests.
Mobile · Storefront · Laravel API · TypeScript system · Demo
Flutter React Laravel TypeScript MySQL Playwright
Status: staging/demo; real SePay Test Mode validation remains a release gate.
An edge-computing system where a Raspberry Pi processes camera input with MediaPipe and OpenCV, publishes driver state over MQTT to an ESP32 alert device, and exposes monitoring data to a Flutter application.
Raspberry Pi ESP32 MediaPipe OpenCV MQTT Flutter
Certificate issuance, revocation, and public verification backed by a Solidity registry. The application connects QR verification, IPFS document storage, MongoDB metadata, and Next.js administration workflows.
Next.js TypeScript Node.js Solidity Hardhat IPFS MongoDB
A reproducible research foundation for a small Vietnamese decoder-only language model, including corpus governance, deterministic data artifacts, Byte-Level BPE tooling, portable compute checks, and human approval gates.
Python PyTorch Byte-Level BPE Data provenance CI
Status: paused, approval-gated research; no language model has been trained.
A full-stack environment for comparing collaborative filtering, matrix factorization, content-based retrieval, and weighted hybrid recommendations, with offline evaluation, a FastAPI service, and a tested React interface.
Python scikit-learn FastAPI React SQLAlchemy Playwright
- Ground AI output in traceable evidence and expose uncertainty honestly.
- Keep API contracts, authorization, money, inventory, and payment state at trusted server boundaries.
- Design retries, offline states, stale responses, and abuse controls as normal operating conditions.
- Treat tests, observability, data provenance, and release limitations as part of the implementation.
Professional contact details are available through the portfolio.
