About Me
I am a computer engineer and AI researcher, now studying for an MS in Artificial Intelligence at the University of Wyoming after a BE in Computer Engineering from Kathmandu University and two years building production software at Pasifikan.
My research sits where model efficiency meets evaluation. I have measured how much factual knowledge a language model can physically hold under low-bit quantization, asked whether vision-language models assembled from the same pretrained parts inherit each other's blind spots, and co-authored a survey on how those models handle negation. I care about experiments that are honest about what they did not show.
On the engineering side I build the systems that research needs and the products that people use: quantization stacks and evaluation harnesses, agentic assistants with retrieval, and a wildlife-vehicle collision alert system now running at the Wyoming Department of Transportation.
