About

About

I’m Rana M Waqas, an AI engineer and engineering leader. I build and ship AI products in industry, and I research the hard problems in LLMs: hallucination, metacognition, and model architecture.

I’ve founded and led engineering teams, and architected AI systems from voice agents to data platforms. That work spans Cisco voice portals at Expertflow, six years at Afiniti on the real-time call-routing AI that served Fortune 100 financial-services and healthcare clients, and founding and leading the engineering team at Ekho, where I shipped an AI voice-simulation platform (Python, FastAPI, LiveKit, OpenAI) that trained 50+ sales reps and cut their onboarding by 30%. I care about products that hold up once real users show up.

Alongside that, I’m completing an MSc in Generative AI at the University of Exeter. My dissertation asks whether language models can recognise the limits of their own knowledge: do LLMs have genuine metacognitive awareness of their knowledge boundaries, or are hallucination-detection methods just exploiting surface-level signals?

This blog is where I work through the engineering: local LLMs and RAG, voice AI, and the unglamorous parts of getting these systems to production.

Get in touch