01
AI-Multiplied Delivery
Claude Code as the primary IDE, sub-agents on isolated worktrees, and a senior reviewer (me) gating every PR. The same engagement that used to need a small team now ships from one operator at 5-10x the cadence.
ExploreAvailableSenior engineering engagements
Recent workSpeed,quality,volume.Getallthree.Finally.
Senior full-stack engineer in Los Angeles. 15 years of React · Next.js · Nuxt taste, now driving a Claude Code agent fleet. The review bar didn't move. The speed, the volume, and the cost did.
Agents in flight10+ in parallel
Throughput20+ PRs / week
ShippingEverAd · IntakeKit · FormForge
One engineer. Ten agents.
The new senior engineer isn't a faster typist. They're an orchestrator. While the industry was deciding whether AI was a gimmick, I built a Claude Code-led pipeline that fans work across parallel agents and lands it in production as small, reviewable PRs. This is what that pipeline actually looks like.
10+Agents in parallelClaude Code instances on isolated worktrees
20+PRs shipped / weekAcross client work and three indie products
~6 minSpec to first PRFrom issue write-up to draft on GitHub
100%AI-assisted outputNo "I can also do it by hand" disclaimer
The pipeline
Brief becomes spec
Every ticket gets a one-page spec written in plain language: goal, constraints, files touched, success criteria. The spec is the contract every agent works against.
Fan-out across agents
Claude Code spawns sub-agents on isolated git worktrees. A research agent reads the codebase. An implementation agent writes the diff. A reviewer agent grades it against the spec before I see it.
Senior review at the gate
I read every PR before it merges. Architecture, security, performance budgets, schema drift, API contracts. Agents are fast. I'm the taste filter and the accountability layer.
CI catches the rest
Typecheck, lint, Playwright on critical flows, Core Web Vitals budget, preview deploy. The bar to merge is the same it always was. Only the cost of getting there collapsed.
The toolchain
Orchestration
- Claude Code (Opus + Sonnet)
- Multi-agent sub-tasks
- Custom slash commands
- Hooks for guardrails
Workspace
- Git worktrees per agent
- Cursor for inline edits
- GitHub PRs as the unit
- Loom for async review
Guardrails
- TypeScript everywhere
- Playwright on critical flows
- Vercel / Cloudflare previews
- CWV budget in CI
Model layer
- Claude API for product features
- Gemini · Imagen · Veo
- Evals before shipping
- Cost & latency dashboards
Metrics are a 90-day rolling average across client and product work. Numbers refreshed roughly quarterly. Want the long version? Book a call and I'll walk you through a recent PR end-to-end.
Projects
PS260
PS260 is a New York-based creative editorial and post-production studio that has been shaping visual storytelling since 2002. The team delivers best-in-class content development, production, editorial, VFX, and motion graphics across every viewable platform, with an award-winning roster spanning JJ Lask, Maury Loeb, Matt Posey, Emile Doucette, Ed Greene, Ashley Ingbretson, BB Blanchard, and the DEARCUT collective, plus dedicated VFX and Sound departments behind campaigns for New Balance, Vogue, and beyond. The site is a kinetic, full-bleed showcase: continuous-scroll editorial reels, dot-paginated section navigation, and a custom video sound toggle, all carefully tuned for performance on a video-heavy frontend backed by Storyblok content and the Simian showcase API.
HudsonArtists
Hudson Artists is a full-service creative editorial and post-production house located in downtown Detroit. Their talented team offers expert editorial, color grading, VFX/finishing (Flame), sound design, and motion graphics. Specializing in commercials and branded content, particularly within the automotive sector (working with clients like Ford and Chevrolet), they partner closely with top agencies and brands. Hudson Artists provides a collaborative environment focused on delivering high-level creative execution. Their beautiful site leverages Simian API integration to showcase their impressive portfolio of work.
NuContext
Nu Context is a dynamic creative production company known for innovative solutions and radical collaboration since 2006. Specializing in turning challenging concepts into reality, their diverse roster of directors and proactive production team excel at elevating campaigns. Explore their portfolio featuring high-profile work for The Oscars, Volvo, The Kardashians, and Dancing with the Stars, showcasing their versatility in producing groundbreaking content.
Ember
Discover Ember's innovative temperature control solutions with ease using our custom-designed mega menu. Optimized for both desktop and mobile devices, this intuitive navigation system provides a seamless user experience. Effortlessly browse our product lines, from the iconic Ember Mug to the advanced Baby Bottle System. Reflecting Ember's dedication to precision engineering and elegant design, our mega menu helps you quickly find the perfect temperature technology for your life.
Process
From scope to ship. A senior engineer at every step.
No agency layers. No handoff drift. You work directly with the person writing the code, integrating the APIs, and answering the pager.
01
Discovery & Technical Scoping
A working session to map the problem. We pressure-test goals against constraints, traffic, content cadence, team capacity, integration surface, and produce a written scope with milestones, success metrics, and a clear architectural recommendation.
02
Architecture & Stack Decisions
Pick the right stack instead of the trendy one. Next.js or Nuxt, headless CMS vs. static, edge vs. serverless, SSR vs. ISR. Every decision documented with trade-offs so you and future engineers know why.
03
Build in Sprints, Ship in Public
Two-week increments, preview deployments on every PR, daily Loom updates if useful. Component libraries, typed APIs, tested integrations, Playwright for the critical flows. You see progress as it lands, not at the end.
04
Performance, A11y, SEO Hardening
Core Web Vitals budget enforced in CI. WCAG 2.2 AA passes. Structured data, sitemap, robots, OG tags audited per template. Synthetic monitoring and analytics instrumentation wired before launch. Not as an afterthought.
05
Launch, Iterate, Measure
Cutover with a rollback plan. Real-user monitoring, conversion instrumentation, and an opinionated dashboard. Then we iterate on what the data says. Not what feels good in the meeting.
Ready to scope a project?
Most engagements start with a free 30-minute call. Bring the problem, leave with a recommended approach and a rough estimate. No pitch deck.
Book a discovery callWhat I do
Senior engineering, multiplied by agents.
I partner with founders, agencies, and in-house marketing teams to ship production React/Next.js and Vue/Nuxt applications. 15 years of taste steering a multi-agent pipeline that delivers in days what used to take weeks.
02
AI Product Engineering
Claude, Gemini, Google Imagen and Veo wired into real product flows. Content generation, conversational intake, scoring and recommendations, hooked to your data with evals, safety guardrails, and cost dashboards.
Explore03
Modern Web Development
Custom marketing sites and web apps in React, Next.js, Vue, and Nuxt. Typed end-to-end, performance-budgeted, and shipped on Vercel or Cloudflare. The same stack I use for a 500+ employee firm and three indie SaaS products.
Explore04
Internal Tools, On Demand
The dashboards, Slackbots, and ad-hoc automations every team wants but never gets budget for. Scoped Monday, shipped Friday. The Jevons-paradox upside of agent-led engineering is that small tools are finally cheap to build.
Explore05
Headless CMS Integration
Storyblok, Sanity, Contentful, and headless WordPress wired into modern frontends. Content models your editors actually understand, with preview, localization, and clean component handoff.
Explore06
Performance & Static Sites
Pre-rendered and edge-rendered architectures, Core Web Vitals remediation, and the kind of audits I've run across 10+ production domains. Sub-second loads, CLS at zero, INP under 200ms.
Explore07
Growth Engineering
UTM attribution, A/B testing with Mutiny and Stellar, landing-page systems, and conversion instrumentation. I close the loop between marketing spend and lead quality with measurable infrastructure.
Explore08
Technical & AI SEO
Schema, structured data, semantic HTML, and Core Web Vitals built in from the first commit. Plus optimization for AI Overviews, ChatGPT, and Perplexity citations. SEO that compounds because it's part of the engineering.
ExploreWhatClientsSay
"
Working with Raffi has been an outstanding experience. His forward-thinking approach to web development elevated our brand's online presence from just a concept video to a fully realized, high-performance site. From the very beginning, he guided us through crucial decisions to like optimal design resolutions and responsive breakpoints across all screen sizes.
Raffi's exceptional communication skills and patience stood out throughout the project to he always took the time to explain complex ideas clearly and made sure we felt confident at every step. He seamlessly led us through everything from server-side setup to the site's launch, leveraging Nuxt.js features like automatic routing, static site generation, server-side rendering, and performance optimization to ensure a fast, scalable, and future-ready web experience. He also integrated a custom CMS that makes updating our content and building new pages simple, even for non-developers.
Raffi's exceptional communication skills and patience stood out throughout the project to he always took the time to explain complex ideas clearly and made sure we felt confident at every step. He seamlessly led us through everything from server-side setup to the site's launch, leveraging Nuxt.js features like automatic routing, static site generation, server-side rendering, and performance optimization to ensure a fast, scalable, and future-ready web experience. He also integrated a custom CMS that makes updating our content and building new pages simple, even for non-developers.
Let's
Connect
Thirty minutes, your hardest problem on the table. I'll show you how an agent-led pipeline ships the kind of work that used to take a five-person team, what the realistic timeline looks like, and which parts are still better off in human hands. No pitch deck. Bring a Notion doc, a Figma file, or just a complaint about your current stack.