The Machine Vision Dispatch San Mateo, California · Est. 2021
Company Profile — Industrial AI

Averroes.ai

The startup that decided the most interesting problem in a chip factory is the defect that gets away - and built a no-code AI to make sure it doesn't.

Averroes.ai logo
The wordmark, set in the company's house black. Named for Ibn Rushd - the 12th-century philosopher who argued that reason and careful observation belong in the same sentence. On a fab floor, that is still the entire job.
Filed under: AI · Manufacturing · Quality Control Reading time ≈ 8 min

Catch Every Defect. Zero Escapes.

Here is a fact about manufacturing that nobody puts on a brochure: every factory, no matter how good, occasionally ships a bad part. The industry even has a word for it - an "escape." A flaw slips past inspection, rides the conveyor to the loading dock, and becomes someone else's problem. Averroes.ai, a roughly fifteen-person startup in San Mateo, has built its entire company around the unglamorous, slightly obsessive goal of making that number zero.

The premise sounds almost too tidy. Manufacturers already inspect their products - often with expensive cameras, microscopes, and, in the semiconductor world, machines that cost more than a house. The problem was never the seeing. It was the deciding. Someone, usually a human, has to stare at thousands of images per shift and judge which speck is a harmless artifact and which is a catastrophic short circuit. Humans are good at this for about an hour, and then they are tired, and then things escape.

Averroes' pitch is that the deciding is a machine-learning problem, and that the machine can be handed to the factory rather than to a data-science team. The platform is no-code: an engineer uploads a modest pile of example images - the company says 20 to 40 per type of defect is enough - labels them, and the model learns what "bad" looks like. No Python, no GPU cluster procurement, no six-month integration. The claim is deployment to production in under five days, which, for anyone who has watched an industrial AI pilot die slowly over three quarters, is the number that actually raises eyebrows.

What makes the small-data claim interesting is that it inverts the usual story about AI. The popular narrative is that these systems are ravenous - that they need millions of labeled examples to do anything useful. But a defect is, in a sense, a rare event. A well-run line does not produce thousands of cracked wafers for the model to study. So a system that demands mountains of defect images is, by definition, useless to the customers who need it most. Averroes designed around the scarcity instead of complaining about it, leaning on techniques like active learning, where the model flags the images it is least sure about and asks a human to weigh in.

Catch every defect. Zero escapes.
— Averroes.ai's tagline, and, effectively, its product spec

That human is a feature, not a bug. Averroes describes its approach as "human-in-the-loop," which is a polite way of saying it did not build the robot that fires the inspector. The inspector stays. The model does the counting and the tedious first-pass triage; the expert validates the edge cases and, in doing so, teaches the model to be better next shift. Continuous learning, in the company's framing, means inspection stops being a thing you install once and becomes a system that adapts as your defects evolve - because they do evolve, as processes drift and new products come online.

The most quietly clever part of the business is what it does not ask you to buy. Averroes sells software, not hardware. It works with the inspection equipment a factory already owns, reading the images those machines already capture and simply interpreting them more aggressively. The company reports that customers catch 40 to 60 percent more submicron defects this way - not by installing better eyes, but by installing a better brain behind the existing ones. For a plant manager whose capital budget is already spoken for, "same cameras, more defects caught" is a rare and welcome sentence.

Then there is the reach. Averroes is deliberately vertical-agnostic, which is a bet that a defect is a defect whether it appears on a silicon wafer, a pharmaceutical tablet, a solar panel, a circuit board, a piece of packaged food, or the inside of an oil pipeline photographed by a drone. Skeptics will note that these are wildly different physical worlds with different tolerances and regulators. Averroes' counter is that the underlying task - look at an image, find the anomaly, classify it, measure it - rhymes across all of them, and that a single engine trained to generalize travels further than a stack of bespoke point solutions.

None of this happens in a vacuum. The AI-inspection space is crowded and getting more so, with rivals ranging from semiconductor-focused startups like SixSense to horizontal players like Landing AI and Instrumental, and incumbents such as Cognex and KLA that have owned the machine-vision aisle for decades. Averroes' differentiation is the combination: no-code, small-data, hardware-agnostic, and fast to deploy. Whether that bundle wins is, as always, a question of execution and trust - manufacturers are conservative for good reasons, and "trust me, ship it" is a hard sell on a line that runs 24 hours a day.

There is a nice symmetry in the name. Averroes - the Latinized name of Ibn Rushd, the 12th-century Andalusian polymath - spent his career arguing that observation and reason are not enemies. A modern inspection line is, if you squint, the same argument rendered in silicon: capture the world carefully, then reason about what you saw. The company would like you to notice that its founder, CEO Tareq Aljaber, spent 23 years shipping products to millions of users at Microsoft, Adobe, Atlassian, and Samsung before pointing that instinct at the factory floor. The lesson he apparently carried over is the least glamorous and most durable one in technology: adoption beats sophistication. A brilliant model nobody can deploy catches exactly zero defects.

99.97%
Reported defect detection accuracy
20-40
Images needed per defect class
<5 days
To production deployment
10x
Faster than manual review

Figures as reported by Averroes.ai. Accuracy and productivity claims are company-stated and vary by application.

Six Ways to Look at a Flaw

Detection

Find It

Submicron-level flaw identification via advanced image processing on existing inspection gear. Reported >98.5% object-detection accuracy.

Classification

Name It

Automated sorting of defect types with a low false-positive rate. Reported >99% classification accuracy.

Segmentation

Map It

Pixel-level outlining of exactly where the defect is. Reported >97.7% segmentation accuracy.

Virtual Metrology

Measure It

AI-driven measurement and process insight pulled from inspection imagery - aimed at semiconductor process control.

Review & Monitoring

Watch It

Human-in-the-loop validation with continuous monitoring, so nothing quietly regresses over time.

Active Learning

Improve It

The model surfaces its least-confident calls for expert review, getting sharper with every correction.

SemiconductorElectronics / PCBPharmaceutical Food & BeverageOil & GasSolar PanelsDrone Inspection

Who, What, Where

Company Vitals

  • Legal NameAverroes.ai Inc.
  • Founded2021
  • HeadquartersSan Mateo, California, USA
  • Team Size~15 employees
  • Business ModelB2B enterprise software (on-prem, cloud & edge)
  • FundingSeed (Dec 2021) — Shorooq Partners, Plug and Play; amount undisclosed
  • IndustryIndustrial AI / machine vision / quality control

The Founders

  • Co-Founder & CEOTareq Aljaber — 23 years of product work at Microsoft, Adobe, Atlassian and Samsung; helped scale Adobe Flash to ~1 billion users; launched Visual Studio App Center at Microsoft.
  • Co-FounderOmar Altamimi

Fun Facts

  • Named after Ibn Rushd (Averroes), the 12th-century polymath of reason and observation.
  • The same engine inspects wafers, pills, solar panels and packaged food.
  • Improves results without buying a single new camera or sensor.

The Paper Trail

Nov 2025
Published whitepapers on EUV defect detection and AI adoption in semiconductor manufacturing.
Jun 2025
Presented at the Vision Spectra Inspection Summit 2025 on how AI is redefining the future of visual inspection.
Dec 2021
Closed a seed round backed by Shorooq Partners and Plug and Play Tech Center.
2021
Founded in San Mateo by Tareq Aljaber and Omar Altamimi.

Quick facts: Averroes.ai

Averroes.ai is a San Mateo-based startup building a no-code AI visual inspection platform that helps manufacturers automatically detect, classify, segment and monitor defects on their production lines. Founded in 2021 by Tareq Aljaber and Omar Altamimi, the company targets semiconductor, electronics, pharmaceutical, food and beverage, and oil and gas producers, promising 99%+ inspection accuracy, deployment in under five days, and models trained on as few as 20-40 images per defect class without writing code or buying new hardware.

Founded
2021
Headquarters
San Mateo, California, United States
Founders
Tareq Aljaber (Co-Founder & CEO), Omar Altamimi (Co-Founder)
Team size
~15 employees
Products
AI Visual Inspection Platform, Defect Detection, Defect Classification, Defect Segmentation, Virtual Metrology
Notable
Reports 99%+ classification, >98.5% detection and >97.7% segmentation accuracy on its platform., Claims models can be trained with as few as 20-40 images per defect class - far below typical deep-learning requirements., Reports customers catching 40-60% more submicron defects on existing inspection equipment and ~30% productivity gains.

Last updated: