Real systems.
Real impact.

A market-intelligence platform for Sarooj Construction (Oman) that continuously scrapes the national tender board, stock-exchange disclosures, and Arabic/English news, then scores every opportunity against the company's own profile: sector, capabilities, geography, and deal size. It maintains an 11-competitor watchlist with head-to-head bid comparisons and an AI-built relationship graph, and writes a weekly executive briefing over ~28,000 historical awards. A parallel multi-agent pipeline runs tender-search, competitor, pricing, supplier, and regulatory analysis, then synthesizes the risk and opportunity view.

Turns a tender into a structured bid decision for Sarooj Construction. It extracts requirements from RFQ documents (PDF/DOCX, with OCR for scans), runs a deterministic bid/no-bid score against the company's capabilities and finances, suggests pricing anchors from a financial twin, and grades subcontractor proposals requirement-by-requirement using retrieval. The model explains each decision but never changes the numbers. Shares one platform and deployment with the market-intelligence module.

An ask-anything analytics assistant over Sarooj's heavy-equipment fleet. Ask in plain language about preventive-maintenance compliance, breakdown and reliability history, repair cost, or asset utilization, and it returns a grounded answer, a table, and an auto-built chart, with every figure computed by validated SQL and the exact query shown for audit. Built for a domain where a wrong number is unacceptable, so it plainly says when nothing matches rather than guess, and refuses to forecast or give buy/sell advice.

An all-in-one project-control platform for a GCC telecom contractor. It reads scanned business documents (purchase orders, invoices, delivery orders, completion certificates) with OCR, extracts the fields, and routes each record through human approval before it counts. From approved records it keeps live per-project financials (invoiced, paid, outstanding, overdue) computed directly in the database, syncs them to Excel, and a separate pipeline reads contracts and scopes of work to draft milestones, tasks, risks, and stakeholders for a human to confirm. The figures come from calculation, not a model guessing over text.

A field-workforce app that runs remote-employee attendance on selfie-plus-GPS punches checked against site geofences, and tracks leave, overtime, meals, shifts, and daily logs. Beyond attendance it carries a full project-management and procurement suite (WBS, Gantt, earned value, cost tracking) and an AI operations copilot that drafts weekly status, flags risks, and answers questions grounded in the live team snapshot. Offline-capable progressive web app with role-based access.

Executive-grade fleet management dashboard for North Kashmir operators. Real-time KPIs, RBAC, Supabase backend (15 tables, ~3K records). Mobile-first command center.

End-to-end AI platform for travel operators: automates itinerary generation, real-time pricing from contracted supplier rates, and multi-channel client communication across WhatsApp, email, and web chat. The system ingests the operator's knowledge base, hotel contracts, tour programs, supplier agreements, and marketing content, to deliver accurate, brand-consistent responses and professionally formatted proposals in seconds. Built-in lead prioritization, SLA monitoring, and an internal dashboard give the team full visibility into every inquiry, while intelligent escalation ensures complex requests reach the right person at the right time.

Evaluates large numbers of CVs together against a job description and ranks the most relevant candidates. Structured extraction, semantic matching, and explainable scoring help recruiters shortlist faster with confidence.

Full-scale DMC platform for an Oman-based tour operator: 16+ pages, 40+ dynamic tour listings, WP Travel Engine + Thawani payments, n8n automation, Arabic RTL, and a custom WordPress plugin (Docker deployment).

AI-powered School Information System for SB Queen's International School. Combines a premium institutional website with a full SIS backbone: student records, admissions, attendance, grading, fee management, and parent/teacher portals. AI layers include an admissions assistant that answers parent queries 24/7, automated report-card generation with personalized teacher comments, attendance anomaly detection, an AI tutor for students, and a principal's dashboard with natural-language queries over school data ('show me all Grade 8 students whose math scores dropped this term'). Role-based access for admins, teachers, parents, and students; Arabic/English bilingual; mobile-first.

Domain-specific AI assistant for telecom operations, internal knowledge retrieval, and workflow automation. RAG architecture with structured document ingestion pipelines.

Real-time stock intelligence across multi-warehouse networks. AI scores risk per SKU, forecasts weekly demand, suggests reorder quantities accounting for lead times, and surfaces critical replenishment alerts before stockouts happen.