He spent two decades shipping tools developers loved. Now he teaches factories to see the defect a human eye would miss.
Averroes was a 12th-century thinker who argued that reason and observation belonged together. Tareq Aljaber put that name on a company that asks machines to do both - not just to see a defect, but to judge it. The choice is not decoration. It is the entire thesis.
Aljaber is the founder and CEO of Averroes.ai, a San Mateo company building no-code AI visual inspection and virtual metrology software for manufacturers. The pitch is deceptively plain: a reliability engineer or quality manager, with no data science training, should be able to upload images, train a model, and deploy it into a production line in hours. The hard part is everything underneath that simplicity.
The problem he is chasing lives at the edge of human vision. In semiconductor fabs, defects can be submicron - smaller than the wavelength of visible light, invisible to a tired inspector squinting at a screen on the third shift. Traditional rule-based systems flag thousands of false rejects, and humans get paid to overrule machines all day. Averroes flips the relationship: the machine learns the inspector's judgment, then applies it consistently, at scale, without fatigue.
What separates the engine, in Aljaber's telling, is that it is data-agnostic. It does not only look at pictures. It ingests time-series signals and tabular process data alongside images, which means it can spot patterns that an image-only system would never catch - a slow drift in a sensor reading that precedes a visible flaw by hours. That multi-modal approach is why the company talks about virtual metrology, not just inspection: predicting a measurement before anyone has measured it.
The arc has a tidy symmetry. Aljaber started his working life in the semiconductor industry - and then spent twenty years away from it, learning how to build and market software the rest of the world actually used. When he came back, he came back as a founder.
Most AI for manufacturing assumes a data science team sitting between the factory and the model. Aljaber's bet is that the team is the bottleneck. The person who knows what a real defect looks like is the inspector, not the PhD. Give that person the tools, and the model gets better and faster.
It is a product manager's instinct more than an engineer's. Across Adobe, Atlassian, and Microsoft, Aljaber's job was to make complicated software feel obvious to the person using it. Averroes is the same move, aimed at a harder audience: reliability and quality engineers who do not have time to learn TensorFlow and should not have to.
The destination he describes is blunt: the future of manufacturing is data-driven, and the company that lets the factory own its own models wins. Whether Averroes is that company is unwritten. But the conviction is not hedged.