AUTOMATION
ENGINEER
Multi-Agent Systems
AI Workflow Automation
Backend Systems
From agent logic
to production API.
A quick map of how I work and what's in the toolkit.
I'm an AI Engineer and Automation Specialist based in Karachi, Pakistan, building for clients worldwide through Upwork. I hold a B.E. in Computer Systems Engineering from Quaid-e-Awam University (CGPA 3.31/4.0).
My work runs the full arc — from n8n workflows connecting Gmail, Airtable, Sheets and WhatsApp, to WhatsApp and Voice AI agents on Vapi, to multi-agent systems built with LangChain and LangGraph.
I also design and optimize production-grade RAG pipelines — parent-child chunking, reranking, vector stores — and pair solid engineering with clear communication, so non-technical clients always understand what's being built and why.
Tools &
technologies.
The stack I reach for across backend, AI, automation, data and DevOps.
Where I've
built this.
Two roles, one thread: turning AI concepts into working systems clients rely on.
- Designed n8n workflows integrating Gmail, Airtable, Google Sheets and WhatsApp, cutting manual work across multiple client automations.
- Built and deployed WhatsApp and Voice AI agents with Vapi, and multi-agent systems using LangChain and LangGraph.
- Built and optimized production RAG pipelines in n8n using parent-child chunking and reranking for better retrieval quality.
- Worked directly with clients to translate technical AI concepts into clear, non-technical language.
- Designed a context-aware LangChain chatbot with conversational memory and tool-based responses.
- Built a document Q&A RAG system with ChromaDB and HuggingFace embeddings, improving retrieval accuracy by 40%.
- Built and deployed RESTful APIs with FastAPI and Flask to serve chatbot and RAG endpoints.
Projects that made
it to production.
A mix of client work and self-directed builds — agents, RAG systems, and automation pipelines.
Customer Support Agent
Modular LangGraph multi-agent system with Triage, Billing, Technical and Complaint agents. A typed SupportState keeps context passing between agents safe and predictable.
EstateGenius
Full-stack real estate RAG chatbot — admins upload property PDFs, users query via conversational AI, with JWT role auth and a persistent Pinecone vector store.
AI Travel Assistant
A ReAct agent that autonomously plans multi-day trips using live weather, nearby places and web search, with session memory and LangSmith tracing for debugging.
B2B Lead Enrichment MVP
Automated lead enrichment pipeline built for a French client — scrapes and matches business data, syncing enriched results straight into their Google Sheets sales workflow.
Zaiqa e Lazzat
Production WhatsApp ordering agent for a food brand — handles menu Q&A, order capture and full conversations through a RAG-backed n8n workflow deployed on Railway.
What clients say.
Sample testimonials — swap in real client quotes as they come in.
Umair rebuilt our order flow as a WhatsApp agent and it just works — fewer missed orders, faster replies, and he explained every step in plain language.
He mapped our lead process into an n8n workflow that used to take hours by hand. Communication was clear and turnaround was quick.
Solid grasp of RAG and agent architecture — he knew exactly where retrieval was breaking down and fixed it without over-engineering the system.
Let's build
something.
Open to freelance projects and full-time roles. Reach out directly or send a message.