You don't need another coder waiting for a perfect spec, or a consultant who hands you a slide deck. You need someone who understands the business and builds the system - the rare overlap. After 15+ years running P&Ls, fundraising and go-to-market, I now build the AI myself, so the bottleneck costing you time and money actually gets found and fixed.
Most AI projects die in the gap between the boardroom and the codebase. I've lived on both sides of it.
I spent 15+ years as a founder, CEO, GM and VP - raising capital, running multinational projects, taking products from idea to market and to exit. Then I went deep into AI and learned to build the systems myself.
So when we talk, I understand your P&L, your customer and your go-to-market - and because an operational bottleneck looks the same whether you're in finance, healthcare or supply chain, I can architect the pipeline and push it to production the same week. No translation layer between what the business needs and what gets built.
Founded & exited an AI industrial-safety startup; co-founded a medtech venture, raised $6M+, tripled valuation through Series A.
GM / VP roles in medtech, agtech & digital health - led $3–5M projects, fundraising rounds, and US / EU / Asia market entry.
MSc Chemical Eng., BSc Materials Eng., BA Physics - Technion. An engineer's rigor behind every system.
13+ live AI systems: multi-agent pipelines, LLM automation, real-time dashboards, data engines - all deployed.
Whether you need one workflow automated or a full intelligent platform - I scope it, build it, and deploy it. Here's where I create the most value.
Orchestrated pipelines where specialized AI agents collaborate on complex workflows - lead handling, analysis, decisioning - delivering in seconds what took hours.
Automate the repetitive, error-prone work: document processing, data entry, outreach, reporting. Reliable systems that run 24/7 without you touching them.
Add GPT, Claude, Gemini or Grok to your product or operation the right way - RAG, vision, function-calling, the correct model for each job, cost-controlled.
Extract, clean and structure data from anywhere - web, PDFs, APIs, competitor catalogs or public records - and pipe it straight into your database or dashboard.
Live, decision-ready dashboards - WebSocket streaming, interactive charts, sub-second data. See what matters the moment it happens.
Dockerized, on your cloud or mine, with SSL, monitoring and CI/CD. You get a working system in production - not a prototype in a repo.
A selection of live, production platforms. Every one started as someone's real bottleneck.

The problem: a flood of unstructured inbound data across 8+ channels, where analyzing each item by hand took hours - so warm opportunities went cold. Nine specialized AI agents now ingest, structure and analyze complex visual and textual data in under 60 seconds each, then surface a ranked shortlist with follow-up messages drafted and ready. Built for real-estate deal flow - but the same pipeline fits any business with high-volume intake to triage.

The problem: generic mass email gets ignored, but genuinely personalized outreach doesn't scale to thousands of recipients. MirageMend uses AI to write context-aware messages for each individual recipient, runs multi-stage drip sequences across rotating inboxes, and detects replies over IMAP - advancing every lead through a Kanban CRM automatically. Outreach that reads one-to-one, at a volume no human team could sustain - whatever you're selling.
The problem: a services business ran its whole sales pipeline out of spreadsheets, email and hand-built PDFs - with no view of which offers were sent, approved or stalled. I built a Kanban proposal platform that carries every offer from draft to completed, auto-generates polished PDFs into the right Google Drive folders, and tracks milestones and payments. Full Hebrew RTL, Google OAuth, dockerized behind SSL - issuing an offer went from an afternoon to two minutes.

The problem: small communities need life-safety alerts - like Home Front Command rocket warnings - to reach every resident the instant they fire, but official apps are easy to miss and no one checks them in time. This bot ingests the alert API in real time and pushes each warning straight into the WhatsApp groups people already watch, 24/7, with zero human in the loop. When seconds matter, the warning is already on everyone's phone - and the same event-to-WhatsApp pipeline fits any critical notification: outages, on-call escalations, logistics events.
4 LLMs in parallel generating trading predictions across 30+ indicators. ~40K lines of Python.
Live options terminal - SignalR streaming, in-browser Black-Scholes, sub-second latency.
Hybrid AI + scraping property analyzer with ROI modeling and Google Maps. Cloud-deployed.
Autonomous agent that screens listing photos with Claude Vision against custom criteria 24/7.
12-stage scraper extracting property data from court records, analyzed with AI, auto-organized.
OpenCV analysis of map imagery - detects properties & water, ranks by proximity.
A live feed of what's happening in AI - auto-updated so you never fall behind.
The things people usually ask before we talk.
Tell me what's slowing you down. I'll tell you honestly whether AI is the right fix - and if it is, how I'd build it. No jargon, no obligation.