Building an Offline-First Knowledge Hub with Obsidian
A Markdown-first workflow for private notes, public articles, research references, and website publishing.
Backend Engineer · System Builder · AI/NLP Learner
I build backend systems with .NET, databases, and practical engineering habits. I also study AI, NLP, and write down what I learn through technical articles, research notes, and project case studies.

Currently
Concurrency, transactions, idempotency, inventory flows, and service boundaries.
C++, algorithms, performance, Linux, and the developer workstation as a system.
Transformers, LoRA/QLoRA, summarization, evaluation, and research workflow.
Featured Articles
Longer notes on backend engineering, databases, Linux, and AI concepts.
A Markdown-first workflow for private notes, public articles, research references, and website publishing.
Why masks use very negative values before softmax, and how padding and causal masks solve different problems.
Designing batch-based stock deduction and reverse stock return with .NET, MySQL, and clear transaction boundaries.
A practical design note for preventing double state transitions when Reject and Cancel requests race each other.
Selected Projects
Each project focuses on the problem, role, decisions, tradeoffs, and lessons learned.
case-study
An e-commerce operations system centered on inventory, order flow, marketplace integration, and backend reliability.
case-study
An e-commerce architecture case study covering frontend, backend APIs, database choices, Docker, Nginx, and deployment.
active
An applied NLP project connecting dataset preparation, model adaptation, evaluation, and engineering workflow.
case-study
A cloud-based logistics and operations platform focused on workflow clarity and integration-heavy backend design.
Research
A paper-review template for tracking method, datasets, results, limitations, and personal takeaways.
A research note for comparing parameter-efficient fine-tuning strategies on Vietnamese summarization datasets.
Recent Notes
A short note on why moving LayerNorm before attention and feed-forward blocks can improve training stability.
A compact note on optimistic concurrency, row versions, and failed updates in EF Core.