Assets
Seven years of production infrastructure. Four business units. 30,000 leads a day. Everything described here is running right now.
Each deep dive below was written after reading every function, every regex, every threshold in the actual production code. Not from summaries. Not from architecture diagrams. From the source files that handle real money and real people every day.
The Four Business Units
1. Apartment Locator
Middleman model. Inbound leads from listing services and paid search. Remarketed to cooperating buildings for 50-100% of first month's rent. Five markets.
Middleman model. Inbound leads from listing services and paid search. Remarketed to cooperating buildings for 50-100% of first month's rent. Five markets.
2. Landlord Rep (Blue Lake / YNY Realty)
I am the leasing office. 35 units across 8 buildings in South Shore, Chatham, and Chicago Heights. First contact through lease signing. 75% commission market-rate, 100% Section 8.
I am the leasing office. 35 units across 8 buildings in South Shore, Chatham, and Chicago Heights. First contact through lease signing. 75% commission market-rate, 100% Section 8.
3. Ken — Renters Insurance
Separate entity (Jamboree Insurance). HO4 renters insurance through a Fortune 500 carrier. SMS-based, outbound-first. Quote, collect payment, bind the policy. Card data never touches my servers.
Separate entity (Jamboree Insurance). HO4 renters insurance through a Fortune 500 carrier. SMS-based, outbound-first. Quote, collect payment, bind the policy. Card data never touches my servers.
4. Financial Analysis
P&L, payment reconciliation, revenue tracking, commission verification. Years of bank statement PDFs across every business entity. The CSV parsers broke, so I read the actual PDFs.
P&L, payment reconciliation, revenue tracking, commission verification. Years of bank statement PDFs across every business entity. The CSV parsers broke, so I read the actual PDFs.
Shared infrastructure: Cloud SQL Postgres, Celery/Redis, GKE cluster (us-central1), SMS, voice AI, Claude Opus 4.6, Gemini (bulk only). Source: github.com/mchopra88/agent-ic
Deep Dives
Each article below is a full technical breakdown — real function signatures, actual thresholds, production regex patterns, and the reasoning behind every decision.
Lead Intelligence
Five signals extracted before the first message sends
Phone area codes, employer extraction, ZIP demographics, financial capacity, credit estimation, background awareness, qualification lanes, confidence cases.
Tonality Engine
How a renter's first text changes everything the agent says next
Renter profiling, five tone maps, style detection, ghost nudge strategies, commitment tracking, over-spam detection, stakeholder awareness.
The Cooperating Network
83,191 buildings, five decay classes, and the forensic pipeline that matches leads to units
Apartment Locator infrastructure. JSONB knowledge store, 7 source lanes, commission verification, matching pipeline, 3-market gap analysis, AI building calls.
The Managed Portfolio
35 units, dual pricing, and the leasing office with zero staff
Landlord Rep infrastructure. 8 buildings in South Shore, Chatham, and Chicago Heights. Section 8 vs market rate. Vacancy management. Unit selection.
The Agent Brain
One loop runs every agent. The LLM decides. Guardrails catch.
Universal 7-step loop, tool execution, Gemini fallback, rate limiting, message guard with 9 blocking categories, duplicate suppression, reasoning traces.
Deployment
Four containers in one pod, Cloud Build, and 16 Discord channels
GKE architecture, container sizing, LoadBalancer config, CI/CD pipeline, environment variables, error logging, SLA monitoring.
Landlord Rep
Operating as the leasing office for 8 buildings with zero staff
Cooperativeness scoring, document pipeline, deal state machine, approval engine, followup cadence, address validation, SMS/call windows.
Ken Insurance
14 intents, one state machine, and the Stripe checkout that binds the policy
Intent classification with exact word lists, 7-status state machine, carrier API flow, payment checkout, followup scheduler, policy generation, database schema.
Financial Analysis
The balance method, and why every CSV parser lied
PDF balance method, 4-tier payment verification, chart of accounts, horizontal truth pattern, collections automation.
By the Numbers
531 commits in 15 months (one person) 1,004,287 lines of Python 7 CLAUDE.md governance files (1,334 lines) 8 deterministic hooks (4 event types) 12 reusable skills 7 slash commands 119 session memory files 4 business units 3 AI agents in production 35 units across 8 buildings (Landlord Rep) 5 lead intelligence signals per inbound lead 14 intent classifiers (Ken Insurance) 5 renter types with distinct tone maps 4-container GKE pod (gunicorn + celery beat + 2 workers) 16 Discord alert channels 30,000 leads/day $0.003 cost per conversation turn 0 human engineers