Josh Rubin - AI Consultant and Product Builder

Josh Rubin

Founder · Builder · AI Developer

Not here to join the herd.
Here to help you build a new one.

AI systems that solve real problems.

Book a Strategy Call

Production System

One agent. Every channel.

Production

Autonomous Multi-Channel AI Agent

One conversation that follows a contact across voice, SMS, WhatsApp, email, and web.

I designed and shipped this as a production system. Five channels resolve to a single identity and one shared memory, so a returning contact is recognized anywhere and the thread picks up where it left off. Every proactive message clears a compliance gate before it sends. I built it for regulated financial outreach, where consent, quiet hours, and hard caps are not optional.

One system. Every channel.Five channels resolve to a single identity and one shared memory.VoiceSMSWhatsAppEmailWebUnifiedMemoryone identity · one threadComplianceGateconsent · quiet hourscaps · kill switchActfails closed

Any channel in, one memory, one compliance gate before anything goes out.

Signal Intelligence

Qualified prospects, found before anyone is contacted.

The system sources from multiple independent data streams and resolves them into a composite signal. No single data point qualifies a prospect. Signals (timing, distress, growth) are cross-referenced across sources, then weighed together so the engine surfaces fit, not noise. A classification layer filters every candidate before it ever enters an outbound channel, so the conversation layer only ever talks to prospects worth talking to.

5

Channels Unified

661

Deals Calibrated

1000+

Tests

How It Works

01

Multi-source signal intelligence. Independent data streams are cross-referenced into a composite signal (timing, distress, growth) that surfaces qualified prospects before contact.

02

First contact goes out on the best-fit channel. That opens the conversation.

03

Their reply lands on any channel and resolves to one identity.

04

Unified memory loads the full history, so the conversation resumes instead of restarting.

05

Work routes to the right tier: a reasoning model where it earns its cost, a classification tier for triage, deterministic code where no model is needed.

06

Every outbound message clears the compliance gate: consent, quiet hours, caps, kill switch. It fails closed.

Shipped Component · Empirical Credit Decisioning

Qualified contacts hand off to a decisioning step that reads structured financial inputs, scores them against an empirical model, and routes the outcome. Most clear automatically. The rest escalate with a full audit trail, so a person reviews the edge cases, not the queue.

Live Demo

Synthetic operations view. Hit the kill switch to pause proactive outreach. Inbound replies are still honored.

Operations · live demo
Autonomous Multi-Channel AI Agent
4,820
Sourced
1,133
Qualified
742
Engaged
318
Decided
Channel activity
Voice
SMS
WhatsApp
Email
Web
one shared memory across all five
Cost · model routing
$209.10 / day
Reasoning tier$186.40
Classification tier$22.70
Deterministic$0.00
right model per task · code where no model is needed
Compliance gate
2,614
allowed
389
blocked
consent state
quiet hours
daily caps
kill switch · fails closed
Live trace feed
waiting for events…
Representative operations view. All data synthetic and anonymized.

Stack

Next.js · Multi-Agent Orchestration · Unified Memory · Real-Time Voice · Compliance Gate

Technical

Composite Signal SourcingCross-Channel Identity ResolutionUnified Conversational MemoryReal-Time Voice AgentsConsent & Compliance GatingTiered Model Routing

Consumer Products

Products that ship.

BashertAI matches screen showing Rachel's profile
BashertAI showing why Yente thinks you'd be great together
BashertAI Yente AI matchmaker conversation
Shipped

BashertAI

AI-powered dating with a matchmaker that actually listens.

Meet Yente, a conversational AI matchmaker that gets to know you through natural dialogue, not checkboxes. She remembers your preferences, understands nuance, and makes introductions that actually make sense.

Technical

Conversational AI AgentStructured Data ExtractionPersistent MemoryContext Management

Stack

Next.js · GPT-4 · Firebase · Stripe

Shipped

LullaPage

Turn any photo into a personalized children's storybook.

Upload a photo, watch it transform into a Pixar-style character, then receive a fully illustrated story with voice-cloned narration. What competitors deliver in 2-3 weeks, LullaPage does in 1-3 minutes.

Technical

Multi-AI OrchestrationImage-to-Character PipelineVoice SynthesisReal-time Generation

Stack

Node.js · Firebase · OpenAI · Gemini · ElevenLabs · Stripe

LullaPage transformation showing a child's photo converted into a Pixar-style character
LullaPage Pixar-style character example
LullaPage Pixar-style character example 2
Golem - AI-powered development companion

Golem

Shipped

An AI-powered app builder. Built to think like a developer, guide like a teacher, and ship like a founder.

Describe what you want in conversation. Watch it build. Preview instantly. Deploy with one click. Two modes: Companion for non-technical users, Developer for full-stack engineers building multi-agent systems.

Conversational Coding

Claude-powered with real-time streaming and auto error correction

Live Preview

Instant browser preview with syntax validation

One-Click Deploy

Publish to custom subdomains instantly

Agent Frameworks

LangGraph, CrewAI, AutoGen, Agency Swarm

An example of the kind of AI systems I build.

Capabilities

What I build.

Multi-Agent Systems

Agents that collaborate, delegate, and converge. I architect multi-agent orchestration from scratch: sequential pipelines, hierarchical routing, swarm intelligence.

  • Sequential Pipelines
  • Hierarchical Delegation
  • Swarm Intelligence
  • Framework Agnostic

Autonomous AI Agents

Custom agents built from the ground up. Tool use, decision loops, memory systems. Not chatbots. Thinking systems that act.

  • Custom Agent Development
  • Tool Integration (Search, APIs, DBs)
  • Decision Loop Architecture
  • Production Patterns

From Concept to Launch

Concept to production: web, mobile, PWA. Full-stack, payments, deployment. Products that make money.

From idea to production in weeks. I move fast and iterate until it's right.

  • Full-Stack Development
  • Payment Integration
  • Cloud Deployment
  • Scale Architecture

Conversational Intelligence

Agents that hold one conversation across every channel. Real-time voice, shared memory, and the guardrails that keep proactive outreach inside the rules.

  • Cross-Channel Unified Memory
  • Real-Time Voice Agents
  • Consent & Compliance Architecture
  • Production-Grade Guardrails

Background

The backstory.

My last company started with a proprietary biotech in a Jerusalem lab, a breakthrough in cannabis bioavailability collecting dust. I saw potential no one else was acting on.

Raised $7.5M. Featured in Forbes. Delivered keynotes.

I build things that ship fast. Today that's AI: conversational agents, voice, text, and image pipelines, products that actually ship. Not decks. Not demos. Working products.

“I work with founders and operators who want something real, not just talk about it.”

$7.5M

Raised

Forbes

Featured

Hopkins

MBA

Production

Systems

Weeks

Not Months

Ready to build something that doesn't exist yet?

30 minutes. No pitch. Let's talk about what you're building.