A focused stack shaped by the kinds of systems I like to own.
Curated for hiring review: not every tool I have touched, but the ones that show up repeatedly in the four systems on this site, from door hardware to LLM agents.
Backend first
I usually start from contracts, jobs, data flow, and operations, then work outward into the product surface.
Shipped systems
Every highlighted tool links to a system it runs in: the MnB access and video platform, the research agent, StageFlow, or AlchemizeCV.
Operator clarity
I value systems that are easier to debug, easier to explain, and less dependent on hidden process knowledge.
Core languages
The languages I reach for most when the work needs to ship and stay maintainable.
Go
coreSite runtimes, agent runtime, orchestrators, CLIs
TypeScript
coreProduct UI, device gateway, browser tooling
SQL
Relational modeling, projections, and operational queries
Backend systems
Durable jobs, typed contracts, and distributed behavior that stays understandable.
NATS JetStream
coreDurable commands and event flow between sites and cloud
Protobuf & JSON Schema
coreContracts generated into Go and TypeScript
AI agent systems
Model-driven work with the guarantees in code: sandboxes, contracts, evidence checks, and a human decision at the end.
Agent runtime (Go)
coreProvider-agnostic loop, persistent shell, resumable sessions
Multi-provider LLM
core10 providers behind 3 API adapters, fallback on exhaustion
Schema-validated output
coreJSON Schema contracts with bounded in-context repair
Evidence gating
coreQuotes checked against captures, a second model scoring citations
Browser tools for agents
coreDaemon-backed real Chrome with a versioned JSON contract
Hardware & protocols
Physical systems where the software is one part of a safety chain.
Product frontend
Frontend work that makes systems usable and legible for technical users and operators.
React / React Router
coreServer-rendered portals and product UIs
Infrastructure, security & delivery
I own the release path and the trust boundaries, not only the application code.
Podman & Quadlet
coreRootless workloads as systemd units
PostgreSQL
coreDurable production state and projected read models
SQLite
coreEmbedded source of truth, outboxes, job queues
Caddy & systemd
Routing, TLS, service management
Signed releases
Digest-pinned images, hash-verified bundles, rollback images
Testing & automation
Tests and automation protect delivery speed; they are not ceremonial coverage.
Playwright
coreBrowser automation, e2e, screenshot and axe checks
Go testing
coreSimulated hardware, real-broker integration, golden runs
Conformance gates
Deployed permissions checked against a hashed manifest in CI
Trivy, gitleaks, govulncheck
Image and dependency scanning in the pipeline