About Me

I’m Laith. This site is where I write about LLMs, evaluation methodology, inference optimisation, and whatever I happen to be reverse-engineering at the moment. The name “lazyevaluator” comes from lazy evaluation in functional programming, the idea of deferring computation until the value is actually needed. It also describes how I tend to think as a ND adult.

By trade I’m an AI researcher and engineer. I spent the last eight years at Nokia as a Bell Labs Distinguished Member of Technical Staff (Bell Labs DMTS 2024, top 0.1% of Nokia technical staff), working across LLM/VLM training, fine-tuning, reward modeling, reasoning, evaluation, inference and AI privacy.

I also worked on generative modelling, applied ML for global telecom operators, and founded Jordan’s first operational research and machine learning consultancy (Holistic Analytics) back in 2010 after returning home with an MSc in Operational Research from the London School of Economics.

I was probably the first Data Scientist in Jordan, but before that name was popularized, I was a statistical analyst conducting multivariate modelling and analysis in the marketing, geolocation, human resources and engineering spaces.

Most of my interest sits in the unglamorous middle of AI research that gets skipped over in louder conversations: curation, evaluation, failure analysis, the slow work of figuring out what a model actually knows.

Evaluation is the most underrated subfield in the stack. If you can’t measure something honestly you don’t really know what you have. That’s been true since I started my career in management systems for industries in Jordan, over 20 years ago.

I’m interested in Neuroscience and Neurodivergence, and am also an aspiring grunge and metal vocalist and guitarist who enjoys hiking in the beautiful outdoors of my country.

Reach me at laith@lazyevaluator.com.