Solutions architect and systems builder working across cloud infrastructure, networking, systems software, and automation.
I turn ambiguous requirements into working systems—from architecture and infrastructure as code through implementation, deployment, testing, troubleshooting, and recovery. My day-to-day stack spans AWS, Linux and Windows, Terraform/OpenTofu, Docker, GitLab CI/CD, Python, Go, C, and C++.
My favorite project and the deepest systems build in my portfolio. AOM-Chess combines bitboard move generation, alpha-beta and quiescence search, transposition tables, classical and NNUE evaluation, Syzygy tablebases, opening books, and multithreaded time-managed search.
The surrounding engineering includes UCI, native ImGui, and WebAssembly front ends; cross-platform CMake builds with scalar, AVX2, and AVX-512 paths; CUDA-assisted NNUE training and dataset tooling; reproducible CuteChess SPRT A/B matches; and a validation suite covering perft, fuzzing, make/undo invariants, protocol behavior, concurrency, endgames, and performance benchmarks.
The source remains private while actively developed.
Native multi-client Windows VPN built in C++23 with Wintun, Winsock, encrypted TCP/UDP transports, Windows routing and NAT integration, a Direct3D/ImGui interface, CMake, Conan, and automated integration testing.
Full-stack Go/Vue network-design and reachability simulator for modeling AWS-style accounts, VPCs, subnets, route tables, Transit Gateway, VPN attachments, customer networks, and policy boundaries before production changes.
Public Go migration of a Tolkien-language dictionary into a source-driven, read-only service with reproducible multi-source imports, bbolt storage, search APIs, data audits, and Docker delivery.
Private Python platform for read-only, multi-account infrastructure inventory, search, historical change analysis, and reporting. It uses least-privilege collection roles, compressed snapshots, retention management, Docker packaging, CloudFormation, GitLab CI/CD, and automated validation.
Upstream fixes accepted into the mainline Linux kernel across tracing, crypto, TCP/IPv6, and netdevsim, with a focus on race safety, callback context, lifecycle cleanup, and maintainer-reviewed correctness.
- C++ language support and hardening for jcodemunch-mcp, including parser behavior, deterministic indexing, incremental storage, documentation, and test coverage.
- Prime-power cutoff tail bounds, an exact-integer verifier and reproducible certificate for a mathematical threshold result.
- Owning difficult problems from design through production recovery
- Reproducible builds, automated validation, and evidence-backed engineering
- Clear boundaries between public proof and private customer or operational details
- Learning whatever the mission requires and carrying the work to completion
Several substantial codebases are private. I describe their architecture, engineering scope, and validation publicly without exposing source, customer data, or sensitive operational details.