Rails hands for the work most teams would rather hand off — performance and scaling, zero-downtime migrations, N+1 hunts, real-time Hotwire without a SPA. Shipped production systems at enterprise scale and helped grow a startup's revenue several-fold. Tested, maintainable code with the decisions written down. Python and Node too when the problem calls for it. Billed corp-to-corp.
Judge the engineer by what he ships. Three live apps you can use right now, plus four open-source Ruby gems & tools — MIT-licensed, on GitHub, and backed by 100 passing tests. Read the code; run the demos.
“GeoGuessr for the ear” — hear real speech, guess the language, place it on a map. 40 languages, a Whisper-verified corpus, served from the edge.
An unsupervised photo-clustering pipeline for heavy equipment — DINOv2 embeddings, UMAP, HDBSCAN, and CLIP labelling, surfaced through an interactive React viewer you can explore in the browser.
An on-device voice wake word for macOS: say “Hey Claude…” and it dispatches a background coding agent. Fully local — openWakeWord + MLX Whisper on Apple Silicon.
Eight production-shaped interfaces in one Rails 8 codebase, no SPA. The flagship, Relay, streams a live Google Gemini response token-by-token to the browser over the Vercel AI SDK protocol — encoded by our own ai_stream gem, with a live wire-inspector and tool-calls. Plus drag-and-drop Kanban synced across browsers, real-time chat, a ⌘K palette, and interactive playgrounds for our open-source gems.
A pure-Ruby encoder for the Vercel AI SDK Data Stream Protocol — the wire format behind useChat/useObject. No Ruby implementation existed; now a Rails backend can drive an AI SDK frontend directly.
A faithful Ruby port of the fzy/fzf fuzzy-matching algorithm — returns a relevance score and the matched character positions for highlighting. Built for command palettes and CLI pickers.
Compiles bash-style glob patterns — *, **, {a,b}, extglobs — into reusable Ruby Regexps. The missing Ruby counterpart to JavaScript's picomatch / minimatch.
Engineering deep-dives on each gem: ai_stream · fzy_score · picoglob
Pick the shape that matches the problem — language-agnostic; we work in your stack, not ours. Each is built around the way real product work actually breaks down, not generic "consulting hours." Scoped up front. No surprise bills.
A drop-in seat on your team, in your stack. We pick up tickets, ship features, review PRs, and carry pager weight — same as a full-time senior, on a contract.
Slow queries, noisy-neighbor problems, a Postgres that's outgrown itself, an S3 bill that's outgrown the company. We diagnose, then fix.
A real REST/GraphQL API for your customers, a third-party data import that has to actually work, an OAuth/SSO story that holds up to security review.
Laravel and Symfony builds, Slim APIs, custom WordPress plugin work, and the harder maintenance work inside older CodeIgniter, CakePHP, and PHP 5/7 applications. The proof is live: one small app built six ways, deployed on six subdomains, plus original PHP packages on GitHub.
War stories and architectural decisions from real Rails work — zero-downtime migrations, N+1 hunts, Sidekiq reliability, real-time Hotwire without a SPA, Postgres at scale — plus the AWS, ML, and AI tooling around it. 24 essays and counting.
Locks, backfills, CONCURRENTLY, and dual-writes — changing a large Postgres schema under a live Rails app without taking it down, and why strong_migrations belongs in every project.
Exactly-once is a lie, so design for at-least-once. Idempotency keys, dedup, retries with backoff, poison jobs, and the dead set — keeping background work correct under failure.
A production Turbo Streams + Stimulus pattern for real-time Rails UI — no React, no API layer, no client-side state to keep in sync. The same approach drives the live demo.
Auditing a decade of Rails code in one afternoon — and the systematic way to find every N+1 before users do.
The AI SDK's useChat speaks a specific SSE wire format; the official backends were JS and Python — Ruby had nothing. The gem that lets a Rails app drive useChat directly.
Every engagement is led by the same senior engineer who answers your first email — someone who has shipped production systems at enterprise scale and built the scaling work that helped grow a startup's revenue several-fold. It's principal-led: when scope demands it, vetted specialists are brought in, and the principal stays accountable for delivery. You get tested, maintainable code, real PRs, decisions written down, and a clean exit when the work is done — the codebase is better and clearer than we found it.
Ruby on Rails is the deepest track and the default — years of senior Rails in production, the gnarly performance and scaling work included. Python and Node are real second languages we reach for when a problem fits them better. The engineering standard — tested, reviewed, maintainable — doesn't change with the language.
The deepest track — years of senior Rails in production, from feature delivery to the gnarly performance and scaling work most teams would rather hand off.
When the work is better in Python, we ship it: FastAPI & Django services, data pipelines & automation, and applied AI/LLM & vision-ML tooling — deployed, not slideware.
Full-stack JavaScript when that's the right tool — typed Node services, React & React Native front ends, and real-time features that have to hold up under load.
Most of our work comes through staffing partners and direct clients who need a senior engineering seat filled cleanly. Here's how that works.
Corp-to-corp through Levelbrook LLC is the default — 1099 and contract-to-hire are fine too. We sign your MSA & SOW or provide ours, carry a mutual NDA, and supply a Certificate of Insurance (General & Professional Liability, $1M / $1M) on request. We onboard through vendor-management systems without friction.
Rates are negotiable and depend on scope and engagement length — corp-to-corp for ongoing work, with fixed-scope projects and longer retainers quoted per engagement. We'd rather understand the problem on a short call than quote blind — you'll have a number within a business day of the brief.
Typically within the week. The trial-week structure means you can see real PRs before committing to a longer term — and walk away at no cost if the fit isn't there.
A principal senior engineer leads every engagement end-to-end — no bait-and-switch to juniors. It's principal-led: when scope demands it, vetted specialists are brought in, and the principal stays accountable for delivery.
Yes — recruiters, staffing partners, and end-clients are all welcome. Send the role or the brief and we'll tell you honestly whether it's a fit.
Recruiters and staffing partners welcome. Direct clients welcome. End-clients welcome. We respond to every brief inside one business day.
Or copy our address: levelbrookteam@gmail.com