DEV Community: Domonique Luchin

Building a Multi-Tenant AI Stack on Supabase + VAPI: How Load Bearing Empire Replaced Twilio and SendGrid with Self-Hosted Infra

Domonique Luchin — Wed, 30 Sep 2026 10:00:08 +0000

When you're running six concurrent businesses from a single technical foundation, third-party SaaS costs compound faster than your margin improves. Load Bearing Empire—a vertically integrated real estate and AI tech operation—solved this by building a self-hosted infrastructure layer that routes 26 VAPI agents through a custom JARVIS Brain Router, backed by Asterisk 18.10 PBX on Vultr and a Supabase project containing 98 normalized tables with 42 production Edge Functions. This article walks through the technical decisions, architecture patterns, and cost-per-transaction math behind eliminating dependency on Twilio, SendGrid, and Make.com automation—replacing them with a unified communication backbone that handles real estate wholesaling, demolition logistics, valet operations, credit repair intake, mineral rights syndication, and SaaS product support from a single API gateway.

Building a Multi-Agent Voice Command Center on Supabase Edge Functions and VAPI

Domonique Luchin — Mon, 28 Sep 2026 10:00:07 +0000

When you're running eight concurrent AI voice agents across residential and commercial properties, you need infrastructure that doesn't break. We built the Quiet Hours Valet Command Center—a production-grade dispatch system handling real-time voice interactions, automation workflows, and client management—entirely on Supabase Edge Functions, VAPI's voice API, and a self-hosted Postgres schema. This article walks through the technical architecture: how we structured an 8-table relational schema to track voice agent states, built five frontend pages on Vercel to manage 8 automation scenarios, and engineered Supabase Edge Functions as webhook handlers for asynchronous VAPI call processing. If you're integrating AI voice APIs into a scaling SaaS or operations platform, this breakdown covers the exact patterns that let us go from prototype to production handling real customer interactions.

Building a Multi-Agent Valet Trash Command Center: Supabase Edge Functions, VAPI, and PostgreSQL at Scale

Domonique Luchin — Sat, 26 Sep 2026 10:00:08 +0000

OPENING PARAGRAPH:

We built a production command center for Quiet Hours Valet that orchestrates 8 concurrent AI voice agents, 8 automation scenarios, and real-time customer management across a vertically integrated real estate and services empire—all on Supabase, VAPI, and self-hosted infrastructure. This article breaks down the architecture: a 5-page Next.js frontend, 8-table Postgres schema for scheduling and routing, Supabase Edge Functions handling VAPI webhook callbacks, and the orchestration logic that lets a single operator manage multiple autonomous agents simultaneously. If you're scaling voice AI beyond proof-of-concept and need webhooks to reliably trigger database mutations, state management without Lambda overhead, and a schema that handles concurrent agent requests, this build is a technical reference.

Building a Multi-Agent Voice Command Center: Supabase Edge Functions + VAPI + Vercel at Scale

Domonique Luchin — Wed, 23 Sep 2026 10:00:10 +0000

OPENING PARAGRAPH:

We built a production voice command center handling 8 concurrent AI agents, 8 automation workflows, and 5 frontend interfaces in 6 weeks using Supabase Edge Functions as our VAPI webhook processor. This technical deep-dive covers our Postgres schema design (8 tables, normalized for agent state management), real-time webhook handling at the Edge, and the architectural decisions that let us scale from a single valet-trash operation to a command center serving Load Bearing's entire portfolio—real estate wholesaling, demolition, credit repair, and mineral rights. If you're building multi-agent systems and wondering how to handle voice webhooks without managing servers, or how to structure your database for complex automation scenarios, this is for you.

Building a Multi-Tenant AI Router for 6 Businesses: Load Bearing Empire's JARVIS Brain on Supabase + VAPI + Asterisk

Domonique Luchin — Tue, 22 Sep 2026 10:00:07 +0000

OPENING PARAGRAPH:

We eliminated $8K/month in SaaS overhead by replacing Twilio, SendGrid, and Make.com with a self-hosted stack built on Supabase, VAPI, and Asterisk 18.10. This article breaks down how Load Bearing Empire architected JARVIS—a central AI Brain Router that orchestrates 26 concurrent VAPI agents across six vertically integrated businesses (real estate, demolition, valet, credit repair, mineral rights, and SaaS)—using PostgreSQL edge functions, a custom PBX layer on Vultr, and deterministic call routing logic. We'll walk through the 98-table schema, the 42 Edge Functions that power real-time agent dispatch, why we chose Supabase over Firebase, and how this approach scales from a single phone to enterprise infrastructure.

Multi-County Real Estate Data Pipeline: Scaling CrawlOS + Supabase for 14-County HCAD Scraping

Domonique Luchin — Mon, 21 Sep 2026 10:00:06 +0000

Building a real estate wholesale operation at scale demands reliable, automated lead generation. At Load Bearing Capital, we integrated CrawlOS into our Supabase infrastructure to scrape Harris County Appraisal District (HCAD) data across all 14 Houston-area counties, generating ~125 scored leads nightly through pg_cron scheduled jobs. This article covers our technical architecture: the 17-table CrawlOS schema design, Actor configuration for the lbc-hcad-actor crawler, Supabase relationship mapping for lead scoring, and production deployment patterns we're using to power our wholesale pipeline. If you're building location-based SaaS, real estate tech, or data-intensive applications on Postgres, you'll find specific schema design decisions and scaling strategies applicable to your own multi-source data ingestion.

Building a Multi-Agent Command Center: How We Orchestrated 8 AI Voice Agents with Supabase Edge Functions and VAPI

Domonique Luchin — Sun, 20 Sep 2026 10:00:06 +0000

OPENING PARAGRAPH:

We just deployed a production command center managing 8 concurrent AI voice agents across a service business with nothing but Supabase Edge Functions, VAPI webhooks, and a carefully architected Postgres schema. This article breaks down the exact technical stack we used to build Quiet Hours Valet's QHV Command Center—a 5-page frontend handling real-time agent state, automation triggers, and webhook callbacks—without a single dedicated backend server. If you're scaling voice AI beyond proof-of-concept and need sub-millisecond webhook processing, deterministic agent routing, and audit-clean transaction logs, this is how we solved it with serverless infrastructure.

Building a Vertically Integrated AI Stack: How Load Bearing Empire Unified 6 Businesses on Supabase, VAPI, and Self-Hosted Aster

Domonique Luchin — Sat, 19 Sep 2026 10:00:06 +0000

When you're running real estate wholesaling, demolition, valet services, credit repair, mineral rights, and SaaS simultaneously—each with different operational requirements—your infrastructure either scales elegantly or collapses under its own weight. Load Bearing Empire chose the former: a self-hosted, open-source foundation built on Supabase, VAPI agent orchestration, Asterisk 18.10 PBX, and custom routing logic that eliminated vendor lock-in with Twilio and SendGrid. This article breaks down how one Black founder architected a unified AI nervous system across 6 distinct business units using 26 VAPI agents, 98 database tables, 42 Edge Functions, and a JARVIS-style Brain Router deployed on Vultr—all running on infrastructure you can actually own and modify.

Building a 14-County Real Estate Data Pipeline with CrawlOS and Supabase: Automating Lead Scoring at Scale

Domonique Luchin — Thu, 17 Sep 2026 10:00:06 +0000

At Load Bearing Capital, we process hundreds of property records across the Houston metropolitan area every night—automatically. This article details how we integrated CrawlOS into our Supabase infrastructure to build a horizontally-scaled web scraper covering 14 counties, generating ~125 qualified wholesale leads per nightly run using pg_cron automation. We'll walk through our schema design (17 normalized tables), the lbc-hcad-actor deployment pattern, and how we structured real estate data pipelines to support both StructCalc AI's structural analysis and Load Bearing Capital's acquisition strategy—all running on self-hosted infrastructure with zero external data vendor dependencies.

Building a Multi-Agent Voice Command Center: Load Bearing Empire's QHV Stack with Supabase, VAPI, and Vercel

Domonique Luchin — Wed, 16 Sep 2026 10:00:06 +0000

OPENING PARAGRAPH:

We built Quiet Hours Valet's command center in 72 hours—a production-grade voice operations platform handling 8 concurrent AI agents orchestrated through VAPI, backed by a Supabase Postgres schema with 8 normalized tables, deployed serverless on Vercel. This article breaks down the architecture: how we structured Supabase Edge Functions to handle VAPI webhook callbacks, designed the agent state machine to route between pickup scheduling, payment processing, and customer notifications, and scaled the frontend from single-page concept to 5 purpose-built pages. If you're building B2B SaaS with voice interfaces or need to coordinate multiple LLM agents through a real-time command layer, this is the technical deep-dive on database design, webhook idempotency, and agent reliability patterns we learned shipping to production.

Building a Multi-Tenant AI Stack for 6 Businesses: How We Replaced Twilio, SendGrid, and Make.com with Supabase, VAPI, and Aster

Domonique Luchin — Sat, 12 Sep 2026 10:00:15 +0000

OPENING PARAGRAPH:

We deployed a vertically integrated AI infrastructure serving 6 revenue-generating businesses—real estate wholesaling, demolition, valet services, credit repair, mineral rights, and structural SaaS—all running on a single Supabase project (ID: whxtjboruayowkqyjvcq) with 98 tables, 42 Edge Functions, and 26 VAPI agents orchestrated through a custom JARVIS Brain Router. This article breaks down the technical architecture: how we eliminated dependency on Twilio and SendGrid by routing all telephony through Asterisk 18.10 on Vultr, replaced Make.com automation with Edge Functions, and built a self-hosted PBX system that handles inbound/outbound calls, IVR logic, and agent dispatch across all six business units. We'll cover Supabase schema design for multi-tenant operations, VAPI agent configuration for industry-specific workflows, and the cost/performance tradeoffs we discovered moving from SaaS platforms to infrastructure-as-code.

Multi-County Real Estate Data Pipeline: Building a 14-County Houston Scraper on Supabase + CrawlOS

Domonique Luchin — Fri, 11 Sep 2026 10:00:09 +0000

OPENING PARAGRAPH:

Real estate wholesaling at scale requires data velocity. This article walks through integrating CrawlOS into a Supabase-backed stack to automate property acquisition across 14 Houston-area counties (Harris, Fort Bend, Montgomery, Galveston, Brazoria, Liberty, Chambers, Hardin, Jefferson, Orange, Polk, San Jacinto, Walker, and Waller). We'll cover schema design (17 tables), the lbc-hcad-actor deployment pattern, pg_cron automation for nightly runs, and lead scoring logic that surfaces ~125 qualified leads per cycle. If you're running a real estate operation on lean infrastructure and need repeatable, autonomous data ingestion at county scale, this deep-dive into CrawlOS + Supabase will show you exactly how to build it.