Senior Backend & Applied AI Engineer focused on agentic developer tooling, production backends, and practical automation.

I build coding-agent systems as bounded, inspectable engineering infrastructure: explicit tools, reusable skills, isolated runtimes, multi-model delegation, and evidence-backed verification.

• Pi/Pipi runtime and workflow tooling
• Agent Skills for code review, browser automation, GitHub planning, and completion verification
• Multi-agent orchestration across Pi, Codex, Claude, Luna, and Terra
• Deterministic web evidence with optional LLM-assisted interpretation
• Backend systems in Go and TypeScript/Node.js
• Production debugging, observability, CI/CD, and Linux automation

An isolated Pi setup with its own pinned runtime and ~/.pipi state. It brings together subagent profiles, parallel workflows, background terminals, browser tooling, Codex-backed tools, reusable skills, and a dark GitHub-style interface without replacing a regular Pi installation. My integration work is maintained in a public fork of davis7dotsh/my-pi-setup.

An Agent Skill and Python CLI that turns structured roadmaps into validated, dependency-aware GitHub backlogs. It supports deterministic batch planning and conflict-safe, idempotent publication of labels, milestones, epics, tasks, checklists, and native sub-issues.

A review skill and contract toolkit that separates confidence from impact, routes only verified serious regressions back as blockers, and keeps closure review focused on the exact remediation surface.

Pi tools for deterministic web search/fetch, Codex-assisted summaries, patch validation, and bounded delegated coding tasks with structured outputs, sandboxes, timeouts, and debug evidence.

A portable Agent Skills package for safe Chrome DevTools automation, with disposable headless sessions for public checks and a separate persistent headed mode for authenticated browser work.

My actively developed fork of MaksBoi/kwispr: Linux voice dictation for Wayland/KDE with cloud, OpenRouter, and local/offline STT; a Rust inference runtime; native KDE shortcuts and tray UI; rootless installation; and optional trusted-LAN inference.

I maintain a public general-purpose skill set used across my Pi/Pipi workflows:

• backend-quality — API, storage, validation, auth, idempotency, and data-safety review;
• frontend-quality — frontend implementation and UI-quality checks;
• devops-quality — configuration, CI, containers, deployment, and runtime readiness;
• visual-composition — product-quality visual hierarchy, responsive composition, states, and interaction polish;
• completion-verification — fresh acceptance evidence before readiness or completion claims;
• git-branching — safe PR-first branches, worktrees, rebases, and synchronization;
• browser-chrome — controlled headed and disposable headless Chrome automation;
• explanatory-html-pages — self-contained technical explainers with readable diagrams;
• modern-skill-revising — focused context and instruction design for modern models.

The standalone code-review and plan-gh-backlog skills extend this set with evidence-driven review and deterministic GitHub backlog publication.

• Obscura upstream commit — merged Rust/CDP contribution attributed to blockedby, covering awaitPromise, DOM event and navigation side effects, click-submit parity, and end-to-end regression tests.
• Go OpenRouter SDK work — fork/contribution work around streaming, reasoning, tool calling, structured outputs, prompt caching, multimodal inputs, and usage fields.
• vibe-practicum-vpn — public-safe Linux/KDE VPN and routing automation, test labs, guarded operations, and redacted diagnostics.
• Agentic Engineering Lab — the broader index of public tooling, workflows, OSS evidence, and sanitized case studies.

I prefer small, sharp PRs with explicit acceptance criteria and fresh evidence:

1. define the problem, constraints, and success conditions;
2. split work into bounded, conflict-aware slices;
3. delegate exploration, implementation, and audit to the right model or tool;
4. keep deterministic evidence separate from model interpretation;
5. verify with focused tests, builds, static checks, browser probes, or API evidence;
6. record what changed, why, and what remains risky.

Previously worked across high-load backend systems, DeFi/Web3 infrastructure, smart contracts, fintech workflows, CI/CD, and team leadership.

Main stack: Go, TypeScript/Node.js, PostgreSQL, Redis, Docker, Kubernetes, Linux, Solidity.

Open to Applied AI Engineer, AI Tooling Engineer, Forward-Deployed Software Engineer, Backend AI Platform, and Founding Engineer roles.