Software engineering · secure systems · applied machine learning
Building systems where trust boundaries, failure modes, and measurable behavior are part of the design.
Based in Singapore. I work across backend and systems engineering, developer infrastructure, practical ML, and security-minded software. I prefer reproducible demonstrations and explicit limitations to claims a repository cannot substantiate.
| Project | Engineering focus | Explore |
|---|---|---|
| RelayFence | Experimental mTLS egress gateway with destination policy, live revocation, and auditable connection handling. Go · network security | Architecture · Verification |
| BranchHarbor | Single-node versioned workspace system using scoped capabilities, content-addressed state, optimistic concurrency, and recovery. Go · backend systems | API · Verification |
| FaultBraid | Experimental three-voter Raft workbench for fault injection, replay, reduction, and inspectable correctness evidence. Go · distributed systems | Architecture · Verification report |
| WattMosaic | Forecasting campus energy demand when sensor readings are incomplete, with explicit missingness patterns and evaluation boundaries. Python · applied ML | Methodology · Notebook |
| Project | Question explored |
|---|---|
| BatchQuay | How do original C++ transformer inference mechanics, request-owned KV pages, and bounded scheduling behave under cancellation and contention? |
| RillScope | How can a query compiler enforce host-defined data access boundaries without trusting user-supplied scope? |
| FacetQuarry | What changes when nearest-neighbor search must respect tenant and metadata filters, with measured recall and work budgets? |
| Driftline · EffectLatch | Exploratory developer tooling and bounded WebAssembly runtime work; refer to each repository for its current maturity and limitations. |
These are engineering experiments, not blanket production-readiness claims. Current support, validation status, and known limitations are documented in the relevant repositories.
01 — Bound authority. Make the trust model, permissions, and operational scope explicit.
02 — Test failure paths. Treat concurrency, malformed input, cancellation, and recovery as first-class behavior.
03 — Show the evidence. Pair implementation with reproducible commands, meaningful tests, and source-bound results.
04 — State the limits. Distinguish completed verification from work still in progress.
Languages: Go · Python · Rust · C++ · JavaScript/TypeScript · SQL
Areas: Backend and distributed systems · Security engineering · Developer infrastructure · Applied machine learning · Data-intensive applications · Full-stack development
My repositories demonstrate different levels of maturity across these areas; the code and evidence tell the detailed story.
LinkedIn · Browse repositories
For reproducible bugs, implementation questions, or documentation feedback, open an issue in the relevant repository.