Real businesses. Real AI agents. Real ROI.

Best Homes Kuwait was losing deals daily. Their sales team took an average of 77 minutes to respond to new WhatsApp leads — by then, prospects had already contacted 3 other agencies. Evening and weekend leads got zero response until the next business day. The team was manually tracking leads in spreadsheets, double-booking viewings, and forgetting to follow up after meetings.

I built an autonomous AI sales agent that runs 24/7 on WhatsApp. The moment a lead sends a message, the AI agent engages in natural Arabic/English conversation, qualifies them (budget, timeline, location, property type), queries the live property database, and sends matching listings with photos and links — all within 60 seconds. It then books viewings directly into the agents' Google Calendar, sends automated reminders 24 hours and 1 hour before each viewing, and follows up after every meeting with new matching listings.

Diagnosing skin diseases from dermoscopic images requires years of specialized training. Existing AI solutions were trained on single datasets, leading to poor generalization across different imaging conditions, patient demographics, and disease distributions. The goal was to build a system that could handle real-world medical imaging diversity — not just perform well on clean, curated data.

I engineered DermaFusion — a Domain-Stratified Ensemble architecture that processes 15 heterogeneous medical imaging datasets simultaneously. The system uses calibrated confidence scoring to deliver clinical-grade predictions, with automatic domain detection that adapts to different imaging conditions. The pipeline includes automated data preprocessing, augmentation strategies specific to medical imaging, and a robust evaluation framework testing across all 15 data distributions.

An e-commerce brand had 3 staff members spending their entire day on repetitive post-purchase operations: manually processing orders from Shopify, updating inventory spreadsheets, emailing suppliers with purchase orders, generating daily/weekly sales reports, and reconciling payments. The process was error-prone — inventory mismatches happened weekly, and supplier notifications were frequently delayed or forgotten.

I built an AI operations agent triggered automatically by Shopify webhooks. When an order comes in, the agent processes the order data, updates inventory across all channels (Shopify + Google Sheets + supplier portal), generates and sends supplier purchase orders via email, creates internal Slack notifications for the team, and compiles automated daily sales summaries and weekly executive reports — all without any human touching the process.

A business owner was spending 3+ hours daily sorting through hundreds of emails — manually categorizing them, identifying urgent items, drafting responses to routine inquiries, and forwarding messages to the right team members. Important emails were buried under newsletters and notifications, causing missed opportunities and delayed responses to clients.

I built an AI email agent using N8N and OpenAI that monitors the inbox in real-time. The agent reads every incoming email, classifies it by urgency (critical / needs response / FYI / spam), drafts personalized responses for routine inquiries, creates task items for action-required emails, and sends a daily digest summarizing everything — organized by priority. The owner now spends just 20 minutes reviewing and approving drafted responses.