About

your brain is running right now on about 20 watts. most of its 86 billion neurons are quiet, and the ones that aren’t fire brief spikes only when something changes. the models we call intelligent do the opposite: they recompute everything, every step, whether anything happened or not. somewhere in that gap is how i think about ai.

i’m a computer engineer and data scientist, currently doing my master’s in data science at montclair state university. most of what i build sits on top of generative ai: agents that read sec filings and get graded against the sec’s own data, a research assistant that’s only allowed to answer from the papers it retrieved, and tools for looking inside transformers and catching models when they quietly drift. generative models can say almost anything. the hard part, and the part i care about, is getting them to say what’s true, show where it came from, and admit when they don’t know.

on the research side i’m exploring spiking neural networks, models that compute with timing and silence instead of dense numbers. i suspect the next leap won’t only come from bigger models but from smarter, cheaper ones that do work only when the world gives them a reason to. long term, i want to build ai you can trust and afford to run: grounded when it speaks, efficient when it listens.