Autonomous agents, RAG, and the infrastructure they run on. Designed, shipped, and operated end to end by one engineer. The kind that files its own work at 3am.
The toolkit · open source
Services
AI consulting, development, and full-stack engineering
Work
Production systems I designed, built, and shipped. From AI platforms to open-source tools
AI Operations Platform
Production · WhatsApp AI live · operated solo since 2025
Watch in theaterTrading AI Analyst
Live · <50ms anomaly alerts · TASE/NYSE arbitrage
Watch in theaterOrchestKit
Open Source · GitHub-starred · 113 skills, 37 agents
Watch in theaterVoice-to-Chart Clinical AI
Healthcare · Hebrew voice→chart · WhatsApp notify
Watch in theaterReal-Estate Connector CRM
Client CRM · WhatsApp + email · AI drafts, human-approved
Open case studyMultilingual Document Intelligence
Production-ready · PT→EN/HE RAG · LangGraph
Open case studyJournalistic Accuracy Scoring
Research · LLM reliability scoring · tiered eval
Open case studyNetwork-Powered Job Discovery
Community · junior-tech jobs · warm-path graph
Open case studyThe Lab
Interactive dashboards and decision boards my AI tooling generates during real work. Not demos: the numbers, issues, and decisions in them are real.
OrchestKit Lab gallery
Every playground is a self-contained HTML artifact from a real session: a 37-issue release-adoption wave, a 98-issue triage board, drag-to-decide standing boards.
Feedback from people I've worked with
“Yonatan built my entire online presence from scratch, portfolio site, SEO, and booking system. He understood my needs as a therapist and created something that truly represents my practice. Professional, fast, and always available.”
Capability × Leverage
A platform engineer's stack compounds. Each capability is a reusable building block. Many are @yonatan-hq/* core packages I wrote once and shipped across every engagement.
Motion showcase · building in public
Real Claude Code sessions, real releases, the making-of rail. They play when you reach them, pause when you leave.
More curated cuts on the version timeline below · tap for sound, expand for theater
The Yonyon identity is drawn from this guy - my husky-mix, the real-life inspiration behind the brand. One amber eye, one ice-blue: the two accent colors this whole site runs on. He supervises every deploy from the couch.
heterochromia, now a design system
George · Chief Morale Officer
Building in Public
Deep-dives into AI agent architectures, RAG optimization, and lessons from shipping production AI systems
The stack · used daily
Every tool here runs this site or the platform behind it, right now. Some links are referral links, your price never changes.
Invoices and receipts
The visual side
Click any tool to open its dossier
Building with AI · WhatsApp
My WhatsApp community: a room for builders, a room for OrchestKit users, and a room for the people bringing AI into their business. Pick yours.
WhatsApp communities · free, no spam
Pinned · The Weekly
Yonyon.AI
One broadcast a week: what got built, what broke, and what's worth your time.
09:00
Yonatan
Builders shipping AI for real: RAG, agents, MCP in production. Stuck on something? Open it here.
Yonatan
Home for OrchestKit users: setups, hooks, agents, and real problems from the field.
Yonatan
Founders, product leads and team leads bringing AI into the business. Decisions and ROI, no code.
About
I build AI that runs itself, then keep it running. One engineer, production-grade: autonomous agents, RAG, custom MCP servers, and the Terraform-managed infrastructure they run on. Solo since 2025, after years building high-scale financial systems.
The proof is my own platform: a smart WhatsApp assistant, a commerce system, content management, and custom MCP servers, all shipped and operated end to end on managed cloud. Like the ops bot that files its own to-do list at 3am.
I also created OrchestKit, an open-source toolkit for working effectively with Claude Code, and I help businesses fold the same autonomous AI into their own workflows.
Core Technologies
Questions
Short, quotable answers about what Yonatan builds, what it costs, and how agents can call this site.
{"messages":[{"role":"user","content":"..."}]} (up to 20 messages, each up to 2000 chars). Public, no auth, rate-limited to 10 requests per IP per minute, backed by the claude-haiku-4-5 model, streaming a text/plain response. See also /llms.txt and the agent-skills index.| Capability | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Backend & Data | ||||||||||
| RAG & Knowledge | ||||||||||
| AI & LLM | ||||||||||
| Multi-Agent | ||||||||||
| Infra & DevOps | ||||||||||
| Design & Frontend | ||||||||||
| Product & Ops | ||||||||||
Tap an engagement or capability to spotlight what it used · tap it again to clear.
15-minute intro call · no commitment