Amna Mahmood

I design and build LLM based systems where retrieval, control, and automation matter more than flashy outputs.

Most AI issues I see are not model problems. They come from weak retrieval, unclear execution logic, or automation chains that fail silently after handoff. My work focuses on building LLM systems that stay reliable once they’re actually used, not just during demos.

I build and refine:

• RAG systems grounded in real documents, databases, or internal knowledge, with attention to chunking, retrieval quality, and failure cases
• LangChain pipelines with controlled prompting, retrieval logic, and predictable behavior
• AI agents that execute multi step tasks using tools and APIs, without looping or acting unpredictably
• OpenAI integrations embedded into backend services or internal tools
• Automation workflows using n8n, Make, or API based orchestration, designed with retries, logging, and clear fallbacks

I work from a systems perspective. Requirements are clarified early, dat