Michél Nguyen.
researcher, engineer,
operator.
I build database engines, ship iOS & macOS apps, publish cybersecurity research, and release music from a single workstation in Berlin. I built most of what is listed below end-to-end.
nowShipping SkeinDB 0.3.17, finishing ShieldLink proof checks, and tuning ShotSift 1.4 for App Store submission. Open full now.txt →
Contents.
All sections of this workstation. Press / at any time to jump.
cd research/
The systems and security work I most want collaborators, reviewers, and technical partners to see first.
SkeinDB
Single-binary database engine with cell-interned MVCC, MySQL compatibility, partial PostgreSQL support, and 20 research tracks shipped inside one executable.
open deep dive → private-cloud securityWardex
Rust-built XDR and SIEM platform for detections, fleet operations, incident response, and hunting without vendor telemetry lock-in.
open deep dive → verifiable storageNexusFS
Offline-first distributed filesystem with signed operations, deterministic replication, and proof-ready storage semantics for edge systems.
open deep dive →ls -la projects/
Active builds across research engines, Python tools, iOS/macOS apps, and games. Click any row to open it.
grep -r research//
Peer-reviewed work, preprints, and the occasional creative publication.
Why some cyber risk in software-intensive firms cannot be priced away
The paper argues that software creates a residual class of uncertainty that does not collapse neatly into ordinary probabilistic risk. Because many security, safety, and compliance questions are undecidable in the general case, even perfect code visibility and abundant classical computation cannot eliminate all uncertainty. The article names that remainder computational Knightian uncertainty (CKU), then connects it to structural opacity in codebases and to practical downstream effects such as incident severity, cyber insurance outcomes, and M&A discounting.
tail -f writing/
Notes, behind-the-paper explainers, working thoughts.
Why SkeinDB ships a single binary
I keep collapsing sidecars back into the database binary because evaluation speed matters more than architectural ceremony at this stage. If the admin plane, runtime evidence, and protocol adapters live in different processes, every claim about the system arrives fragmented. A single executable forces the product, the research surface, and the operational story to line up in one place, which makes the project easier to inspect and much harder to hand-wave.
play music/
Six releases out in 2025. Lo-fi, DnB, and field beats from Da Nang and Da Lat. Click a row to preview.
cat now.txt/
What I'm actively working on. Updated weekly.
This week's bench
- monskeindb 0.3.17 — finishing the MVCC reclamation pass + writing the engine-handbook chapter on it.
- tue–wedshieldlink in press at IJRCOM — final metadata and proof checks.
- thupre-doctoral — drafting malware classification lit review, drift under adversarial relabeling.
- frishotsift 1.4 — burst-grouping perf regression on iPhone 12 mini, then App Store submission.
- weekendMixing a new DnB track. Reading Designing Data-Intensive Applications again.
which tools/
What I reach for weekly. Not exhaustive.
systems & db
- rustdaily
- c / c++weekly
- mvcccore
- storage enginescore
- mysql / postgrescore
- query planningcore
apple
- swiftdaily
- swiftuidaily
- spritekitweekly
- widgetkitcore
- cloudkitcore
- visioncore
protocols
- http/3 · quiccore
- axumweekly
- json-rpccore
- restcore
- node.jscore
- wasm sandboxingcore
research
- federated learningactive
- differential privacyactive
- intrusion detectionactive
- formal verificationcore
- theoretical physicsinterest
- academic writingdaily
history -e/
Formal study, peer-review work, ongoing pre-doctoral preparation.
Pre-doctoral Research Preparation — Cybersecurity (MPhil/PhD)
General Sir John Kotelawala Defence University, Sri Lanka. Research preparation phase for graduate study.
- Research synopsis + topic development on malware classification with ML.
- Literature review, static/dynamic feature engineering, robustness/drift evaluation.
- Outputs in flight: preprint, public GitHub artifacts, curated dataset prep.
B.Sc. Computer Science
University of the People · Pasadena, California, US.
A.S. Computer Science — High Honors
University of the People · Pasadena, California, US.
Reviewer — 6th Student Symposium, Faculty of Computing
General Sir John Kotelawala Defence University, Colombo, Sri Lanka.
ping minh/
Research, partnerships, app questions, or speaking: pick a channel.