AI Integrator & Automation Architect

I build systems that handle the work businesses are still doing by hand.

Law firms chasing client intake. E-commerce brands checking payments manually. Clinics booking appointments through phone calls. Real estate teams following up on leads one by one. The bottleneck is almost always the same — too much manual work, not enough automation.

⚖️ Law Firms  ·  🏠 Real Estate  ·  🏥 Clinics  ·  🛍️ E-commerce  ·  💼 Agencies  ·  🚀 SaaS

An autonomous sales agent for Facebook Messenger that handles the full sales cycle — product questions via text, voice, or photo — through to inventory checking, fraud detection, and order placement. Transfers to a human agent when the situation needs it.

Stack: n8n · OpenAI · Google Gemini · Supabase · Pinecone · Postgres

flowchart TD
  User((User Input)) --> Webhook[n8n Webhook]
  Webhook --> Router{Input Type?}

  Router -->|Image| Vision[Gemini Vision]
  Router -->|Voice| Audio[Whisper Transcribe]
  Router -->|Text| Token[Token Check]

  Vision --> Agent
  Audio --> Agent
  Token --> Agent

  Agent["🤖 AI Sales Agent (RAG)"]

  subgraph Tools [Agent Tools & Memory]
    direction TB
    Agent <--> GSheets[("Google Sheets: Stock/Orders")]
    Agent <--> Vector[Pinecone: FAQ/RAG]
    Agent <--> Fraud[BD Courier API: Fraud Check]
    Agent <--> Handover[Supabase: Human Handoff]
    Agent <--> Memory[Postgres Chat Memory]
  end

  Agent --> Messenger[FB Messenger Reply]

  style Agent fill:#FF6D5A,color:#fff,stroke:#fff
  style Tools fill:#2D3B45,color:#fff

Key Features:

• Gemini Vision identifies products from customer photos
• OpenAI Whisper transcribes and understands voice notes in real time
• Real-time inventory sync with Google Sheets — checks stock before confirming
• BD Courier API fraud scoring on COD orders before acceptance
• Pinecone vector DB for FAQ and policy retrieval
• Postgres memory keeps conversation context across sessions
• Supabase-triggered human handoff when complex support is needed

Repo: AI-Sales-Agent-Omni-Channel-Automation

A full order management system for businesses selling through Instagram or Facebook DMs. No Shopify. No website. Customers submit orders through a form, payment screenshots upload to Google Drive automatically, fraud detection runs, and the owner approves or rejects from Telegram in one tap.

Results: 40hrs/week → 2hrs  ·  12 fake orders/week → 0  ·  25,430+ orders processed  ·  99.1% verification rate  ·  <1.5s response time

Stack: n8n · Airtable · Google Drive · Google Sheets · Gmail · Telegram Bot

graph TD
  A[Customer Order Form] --> B[Validate & Normalize Order]
  B --> C[Payment Screenshot → Google Drive]
  C --> D[Attach Screenshot URL]
  D --> E[Duplicate Data Check]
  E --> F[Fraud Detection Logic]
  F --> G{Duplicate or Fraud?}

  G -->|True| H[Send Fraud Alert via Telegram]
  G -->|False| I[Log Order to Google Sheets]
  I --> J[Send Order Placed - Gmail + Telegram]

  K[Telegram Callback] --> L[Parse Callback Data]
  L --> M[Fetch Order from Sheet]
  M --> N[Merge Order Data]
  N --> O{Owner Decision}

  O -->|Approve| P[Status: Confirmed]
  O -->|Reject| Q[Status: Rejected]
  O -->|Fraud| R[Status: Fraud]

  P --> P1[Send Confirmation - Gmail + Telegram]
  Q --> Q1[Send Rejection - Gmail + Telegram]
  R --> R1[Send Fraud Message - Gmail + Telegram]

  style G fill:#f9a,stroke:#333
  style O fill:#f9a,stroke:#333

Repo: From-DM-to-Delivered-Zero-Website-Order-Flow

An AI agent that retrieves contextually relevant answers from a custom database using vector search. Built for businesses that need a chatbot trained on their own data — not generic internet knowledge.

Stack: n8n · Pinecone · Groq · MongoDB

graph TD
  User((User Query)) --> Hook[n8n Webhook]
  Hook --> Embed[Generate Embedding]
  Embed --> Search{Vector Search}
  Search -- Query --> PDB[(Pinecone / MongoDB)]
  PDB -- Retrieved Context --> AI[AI Language Model]
  AI --> Result[Synthesized Answer]
  Result --> Response[Return to User]

  style Search fill:#f9f,stroke:#333
  style PDB fill:#27272e,stroke:#fff,color:#fff

Key Features:

• Intent recognition before retrieval
• Dynamic context injection based on query similarity
• Low-latency responses via Groq inference
• Works with any structured or unstructured business data

Use Case: Custom AI chatbots for law firms, clinics, and agencies that need to answer questions from their own documents and policies.

An OTP verification layer for WooCommerce that blocks fake orders before they reach fulfillment. Verifies customer phone numbers at checkout via automated SMS, adds a confirmation step that most fraud bots and fake buyers won't complete.

Stack: n8n · WooCommerce · SMS API · Google Sheets

graph LR
  A[Customer Places Order] --> B[n8n Webhook Triggered]
  B --> C[Generate OTP Code]
  C --> D[Send SMS to Customer Phone]
  D --> E{OTP Verified?}
  E -->|Yes| F[Order Confirmed in WooCommerce]
  E -->|No / Timeout| G[Order Cancelled + Alert]
  F --> H[Log to Google Sheets]

  style E fill:#f9a,stroke:#333

Repo: Automated-WooCommerce-OTP-Fraud-Prevention-System

An automated pipeline that processes logistics data, validates it against business rules, flags errors, and generates structured reports. Removes the manual spreadsheet work from operations and compliance teams.

Stack: n8n · Google Sheets · AI Model · PDF/Doc generation

graph LR
  A[Raw Logistics Data] --> B[n8n Trigger]
  B --> C[Data Validation Layer]
  C --> D{Passes Rules?}
  D -->|Yes| E[AI Report Generation]
  D -->|No| F[Flag Errors + Notify Team]
  E --> G[Export PDF / Doc]
  G --> H[Send to Stakeholders]

  style D fill:#f9a,stroke:#333

Repo: AI-Logistics-Report-Documentation-Validation

An automation pipeline that takes raw product photos and turns them into UGC-style ad content using AI vision and copy generation. Built for marketing agencies producing high volumes of creative assets.

Stack: n8n · Google Gemini · Cloudflare

graph LR
  A[Raw Product Photo] --> B[Gemini Vision Analysis]
  B --> C[AI Ad Copy Generation]
  C --> D[Asset Design & Formatting]
  D --> E[UGC-Ready Ad Asset]

  style B fill:#8E75B2,color:#fff

Live: nbgc.pages.dev

An AI-powered studio that manages the full content workflow for solo creators — planning, scheduling, and distribution in one automated system.

Stack: n8n · Supabase · Lovable

graph LR
  A[Content Idea Input] --> B[AI Content Planner]
  B --> C[Schedule & Queue]
  C --> D[Auto Distribute]
  D --> E[Facebook]
  D --> F[Instagram]
  D --> G[Other Platforms]

  style B fill:#3ECF8E,color:#fff

Live: ureatorflow.pages.dev

A consultation form that sends personalized reply emails automatically. When a client submits their request, an AI Agent reads it, writes a tailored email based on their specific service interest, and Gmail delivers it within seconds. No manual follow-up needed.

Stack: n8n · Google Gemini · Google Sheets · Gmail · Structured Output Parser

graph LR
  A[Client Submits Form] --> B[Log to Google Sheets]
  B --> C[AI Agent Reads Request]
  C --> D[Structured Output Parser]
  D --> E[Subject + Body Extracted]
  E --> F[Gmail Sends Personalized Email]

  style C fill:#8E75B2,color:#fff

Repo: AI-Vibe-Client-Intake-Automation

• 🔨 Building: UreatorFlow — AI operating studio for solo creators  ·  NBGC — product photos to UGC ads
• 📚 Learning: Multi-Agent Orchestration and scaling SaaS with n8n + Supabase
• 🤝 Open to: Freelance projects and open-source automation collaboration
• 💬 Ask me about: n8n, AI Agents, Supabase, CRM integrations, Telegram bots

Category	Tools
Automation
AI & LLM
AI Frameworks
CRM & Comms
Database
Web Dev
Dev Tools

Solo engineer. Select projects. Deep work.