Software Engineer · Systems Architect
Systems · Networks · Scale · Reliability · Security · Observability
MCP Trace · Open Source · Engineering Surface
architecture > isolated features · evidence > assumptions · reliability > hype
I design software as systems, not isolated features.
My work centers on failure-aware architecture, runtime behavior, networking, observability, security, and performance. I prefer explicit invariants, reproducible tests, narrow interfaces, measurable behavior, and telemetry that remains useful when a system is under pressure.
Most commercial work is private. The public work below is selected engineering evidence.
A security-conscious observability gateway for the Model Context Protocol over Streamable HTTP.
MCP client ──────▶ MCP Trace ──────▶ upstream MCP server
│
├── Prometheus metrics
├── OpenTelemetry traces
└── sanitized NDJSON recording + controlled replay
- Transparent JSON and SSE proxying without taking protocol authority away from the upstream server.
- Prometheus-compatible metrics, OpenTelemetry spans, W3C Trace Context, structured logs, and MCP-aware inspection.
- Metadata-only recording by default, explicit body capture, credential redaction, hashed legacy session identifiers, and owner-only recording permissions.
- Replay is dry-run-first with explicit execution, concurrency, rate, and timeout controls.
- Compatibility across modern 2026 Streamable HTTP and the 2025 MCP transport revisions.
Architecture · Security model · Recording schema · Releases
Current upstream patches under maintainer review:
| Project | Contribution | Engineering scope |
|---|---|---|
NASA F´ · nasa/fprime |
#5670 — parameter validation macro | C++ framework API; centralized parameter-validity semantics; refactor across 9 files; four-state coverage; 129/129 unit-test targets |
Google gVisor · google/gvisor |
#14097 — PTY positional-write semantics | Linux VFS/devpts behavior; ESPIPE compatibility for positional PTY writes; syscall and filesystem regression coverage |
Cloudflare Workers SDK · cloudflare/workers-sdk |
#15158 — Workflow batch deletion results | Runtime behavior; not-found/failure propagation; duplicate-safe result mapping; regression coverage across success and failure paths |
OpenAI Cookbook · openai/openai-cookbook |
#2495 — Windows path portability | Cross-platform repository integrity; Windows-compatible checkout; tested path-portability guard across 3,070 tracked paths |
I am deliberately focusing on contributions where the patch exposes real engineering judgment: compatibility semantics, failure behavior, tests, invariants, portability, and maintainability rather than drive-by cosmetic changes.
| Layer | Working set |
|---|---|
| Systems & runtime | C, C++, Rust, Go, Linux, VFS/filesystem semantics, process and network behavior |
| Services & infrastructure | TypeScript, Node.js, REST, WebSockets, PostgreSQL, Redis, Docker, CI/CD |
| Observability | OpenTelemetry, Prometheus, structured logging, tracing, recording, replay, operational diagnostics |
| Application engineering | React, Next.js, realtime interfaces, performance-sensitive frontend systems |
| Security & delivery | threat-aware design, input validation, access-control boundaries, CodeQL, secret scanning, GitHub Actions |
| Automation | Python, Bash/Shell, PowerShell, repository tooling, validation and release workflows |
- Architecture before implementation.
- Design for failure, not ideal conditions.
- Treat security, performance, and observability as initial constraints rather than cleanup work.
- Prefer behavior that can be tested, measured, reproduced, and explained.
- Keep interfaces narrow and operational failure modes explicit.
- Optimize for systems that remain understandable after the original implementation context is gone.
runtime systems · distributed infrastructure · networking · observability tooling · performance · open-source engineering
The objective is simple: build systems whose correctness is visible in their behavior, tests, architecture, and failure handling—not in the number of technologies listed beside them.