Muhammad H.

I'm an AI engineer who builds production AI platforms end to end: the product, the AI agents and retrieval inside it, and the security and data architecture underneath. Founding AI engineer at a fintech, AI lead on a multi-tenant compliance platform, and builder of AI agents that have automated 8,700+ real support tickets. PLATFORMS I HAVE BUILT Fintech investing platform (founding AI engineer) - AI research agent with guardrails and a streaming safety filter - MCP server with hashed, revocable access tokens - Bank-data security: two-step sign-in, consent records, rate limits, monitoring Multi-tenant AI compliance platform (AI lead) - Bilingual English/Arabic GRC platform with 16 specialised AI personas - Hybrid retrieval with reranking and per-tenant access control at retrieval - Graph memory, browser automation and RAG evals (RAGAS, Arize Phoenix) Compliance operations platform for a school - Full-stack Next.js app with role-based access for 7 staff roles - FERPA-aware data handling and compliance monitoring - CRM sync and scheduled jobs Bank-feed financial module inside a live field-service SaaS - Plaid bank and card feeds combined with CRM revenue - Deterministic calculation engine, self-service setup and an owner-only view - Encrypted tokens, signed webhooks, per-company isolation, no AI on bank data PRODUCTION AI SYSTEMS - AI customer support agent for a UK broadband provider: 8,700+ tickets automated and 79,000+ agent minutes saved. It checks the CRM and live outages mid-chat and routes escalations across 7 departments. - WhatsApp retention engine for the same provider: consent-gated personalised offers with SMS fallback, AI reply handling and a ticketed handover to the retention team. - Email and web-form intake for a translation company: a 224-node n8n workflow, self-hosted in Docker, with schema-locked AI steps and a non-AI gate that escalates to a human. - Change detection across tens of thousands of directory listings, with two-stage confirmation and an append-only history. - Native iOS and Android launch of a marketplace app, approved on the first App Store submission. HOW I WORK I build AI-native with Claude Code, with small reviewed pull requests, automated tests and CI gates on every change, so speed never costs reliability. Architecture comes first: tenant isolation and row-level security, strict output schemas on every AI step, evals, human approval where a mistake costs money, and monitoring in production. STACK AI: Claude and OpenAI APIs, agents, MCP, RAG (hybrid retrieval, reranking, pgvector), evals. Product: Next.js, React, TypeScript, Python, FastAPI, .NET, Supabase and Postgres, Vercel, Docker. Integrations and automation: Plaid, Microsoft Graph, Google APIs, CRMs, n8n, Make, Zapier, Power Automate. Voice and messaging: Retell, Vapi, ElevenLabs, WhatsApp Cloud API, Twilio. Engineering: Claude Code, GitHub Actions, Playwright, Sentry. CRMs: Jobber, Hubspot, GoHighLevel, Airtable, Sheets, TitanService. INDUSTRIES Fintech, compliance and GRC, telecom, field service, education, translation, real estate, insurance. If you are building an AI product, or need AI inside one you already run, send me the brief. I will reply with the architecture I would use and a fixed-price first milestone.