Muhammad Moazzam Q.

I take AI products from prototype to production: AI agents, voice agents, document intelligence and the full-stack SaaS system around them. For the last 12 months I've been Lead Full-Stack AI Engineer on an enterprise property management platform: 1,700+ hours on a production AI system used by real customers, leading a team of 4 engineers and owning DevOps. I also built the customer service voice agent for a $10M/year DTC ecommerce brand, handling live cancellation, returns, onboarding and general enquiry calls against real-time order data, company policies and the CS team's own scripts. ───────────────────────── ◆ WHY THIS MATTERS TO YOU ───────────────────────── Most AI developers build demos. I build systems that run on real data, with real customers, permissions and business rules, in two demanding domains: property management and real estate, and high-volume DTC ecommerce customer service. Problems I've actually worked through: ▸ A tenant calls from an unrecognized number and the system has to work out who they are, which tenancy they belong to, and which property they are calling about. ▸ A maintenance call needs more than intent classification. The system has to determine what happened, whether it's an emergency, who is responsible, and what happens next. ▸ A new buyer enquiry is worth the most in the first few minutes. It needs to be qualified, matched to the right listing and agent, and booked for a viewing before it goes cold. ▸ A certificate is not just a PDF in a database. Gas safety, EPC and other compliance documents contain dates that trigger work, notifications and follow-ups. ▸ Rent arrears are not a "send reminder" workflow. Timing, tenancy status and previous communication all matter. ▸ One property can involve a tenant, landlord, buyer, letting agent, property manager and contractor, each with different permissions and actions. ▸ A customer calls to cancel, but the order may already be picking, packed or shipped. The agent has to check live order status and apply the right policy: cancel, intercept, or convert to a return. ▸ A return request isn't a yes/no. Purchase date, item condition, order history and policy exceptions decide the outcome, and the agent has to explain it the way your CS team would. The same patterns apply to any SaaS where AI has to act on real business data. → If you're in property, real estate or ecommerce, you won't pay me to learn your business from scratch. ───────────────────────── ◆ AI VOICE AGENTS ───────────────────────── Property management: ▸ Tenant, landlord and letting enquiries ▸ Caller and tenancy/property matching ▸ Maintenance triage, emergency vs routine ▸ Job creation, follow-ups and rent chasing Real estate: ▸ Buyer, seller and renter lead qualification ▸ Speed-to-lead callbacks ▸ Listing questions from live property data ▸ Viewing scheduling and CRM updates Ecommerce: ▸ Cancellations, returns and exchanges against live order data ▸ Policy-driven decisions, not scripted guesses ▸ Onboarding and general enquiries ▸ Your CS team's scripts, tone and escalation rules Every agent includes transcription, summaries and human escalation, and performs actions in the underlying system rather than simply taking a message. ───────────────────────── ◆ DOCUMENT INTELLIGENCE ───────────────────────── ▸ Certificates, EPCs, tenancy documents, invoices and statements ▸ Offers, purchase agreements, disclosures and listing documents ▸ Extraction of prices, parties, key dates and conditions ▸ Deadline tracking, document generation and e-signatures Example: a gas safety certificate is uploaded → matched to the property → processed → compliance record updated → the right workflow triggered. ───────────────────────── ◆ MCP + AI WORKFLOWS ───────────────────────── I build custom MCP servers giving AI agents governed access to live properties, tenancies, leads, orders, financial records, maintenance tasks, compliance certificates and documents, plus the workflow engine around calls, messages, uploads, status changes, key dates and no-response timeouts. The AI handles language and unstructured information. The application handles state, rules and execution. ───────────────────────── ◆ PRODUCTION ENGINEERING ───────────────────────── ✓ Multi-tenant architecture, data isolation, RBAC and audit trails ✓ GDPR-aware data handling and consent ✓ Azure Key Vault, managed identities, encryption at rest and in transit ✓ Usage metering and tenant accounting ✓ Integration/E2E testing, CI/CD, monitoring and production support Stack: Python, FastAPI, LangGraph, LangChain, .NET 8, Azure, PostgreSQL, Docker Integrations: AppFolio, Twilio, Vapi, Retell, ElevenLabs, OpenAI, Claude → Send me what you're building and I'll give you a straightforward answer on scope, approach and timeline.