Wahab A.

I've earned $100K+ on Upwork building AI systems that actually reach production. Right now I'm deploying a self-hosted executive agent on Gemma and vLLM for a client who needs the model running on their own infrastructure.

Wiring up a chatbot is easy. Making one reliable is the hard part. It has to call the right tools, hold its state together when a job fails halfway, and still give the right answer once real users and messy data show up. That's the work clients bring me.

⚙️ AI Agents & Automation
Python agents built with LangChain, LlamaIndex and CrewAI. n8n workflows that handle retries and failures cleanly instead of breaking silently. Claude and Claude Code automation, plus MCP and tool integrations across the OpenAI and Anthropic APIs.

📚 Production RAG
Retrieval pipelines, chunking and embeddings, answers grounded with citations, and hardening live systems against prompt injection and data leakage. My longest contract here is 1,792+ hours and $80,000+, building production RAG and Azure OpenAI integration for one client, and it's still going.

🔒 Self-Hosted & Open-Weight LLMs
When client data can't leave their environment, I deploy and serve private models like Gemma on vLLM.

📊 Applied ML
Patient risk scoring, recommendation engines, predictive analytics on real business data.

I've done this for Thoughtful AI, Modulate AI and Sieve, and worked as a full-stack Python engineer on an Upwork Enterprise contract. A good number of clients have come back to hire me again.

How I work is simple. I scope it before I build it, break the edge cases before delivery, and leave you something documented your team can run without me.

If you've got an agent to build or a RAG pipeline giving wrong answers, send me the details and I'll come back with a plan.