Hacarus: Combining Small & Good Data Centric AI with Robotics for Smarter Visual Inspections

The Silicon Review

With access to centuries-old craftsmanship cultures that collide with immediate technological progress, Hacarus, a quiet name of the city of Kyoto, has become one of the most influential AI companies in Asia. Though less dazzling than its Silicon Valley brethren, this Japanese company is pioneering the use of sparse modeling technology, which is radically changing industrial automation in the manufacturing epicenters of the continent.

The Kyoto Difference: When Less Data Means More Insight

When most AI businesses are scurrying after bigger datasets, the Hacarus has done just the opposite. Their state-of-the art sparse modeling technology is proving superior to deep learning in an area where it performs poorly - accurately deriving insightful knowledge based on small datasets. Such anti-intuitive technique has been of special use to:

Such industries as semiconductor manufacturing and aerospace engineering where defects are very few and yet have the potential of being disastrous. Conventional AI systems presumably have to be fed thousands of faulty samples to discern trustily - and these merely do not exist when it comes to superior manufacturing. In only dozens of examples, Hacarus technology can attain the 99.9% accuracy.

Fast manufacturing processes whereby product lines are changing constantly. Lighter models might take hours as opposed to weeks of retraining a system with new data on the part of competitors. The versatility has enabled them to become the most vital of all materials to the manufacturing companies of electronics which operates on a context of continual changeover of the products.

Strictly regulated industries in which complete auditability is needed. The AI proposed by Hacarus can explain each decision made, unlike the black box approaches of deep learning where it is hard to see what caused an AI model to classify a defect. This has been vital to the medical device makers as well as car component suppliers who are subject to stringent compliance obligations.

Its visual inspection systems are being used in 12 countries to inspect everything including Shinkansen bullet train parts and precision surgical tools. In highly optimised deployments, they reach the far end of the spectrum by detecting with 99.97 accuracy with 90% fewer training data than deep learning solutions.

Industrial AI That Speaks Human

The human-centric strategy of artificial intelligence is the best feature of Hacarus. The company has come up with tools that enhance human expertise instead of looking at AI tools as a replacement to skilled workers.

The Check Series perhaps is the closest embodiment to this philosophy. Integration of collaborative robots and interpretable AI vision systems, will offer 360-degrees part inspection with complete transparency. The company Asahi Kosei experienced a 66 percent drop in manual inspection after integrating the system into their precision components division, in fact improving their defect defect rates.

Hacarus has revolutionized the compulsory Kiken Yochi (hazard prediction) in construction industry in Japan by their Workplace Safety KY system. They have assisted large construction contractors to cut on-site accidents by 40 percent by transforming what was once a paperwork exercise into a smart AI enabled workflow. Its capacity to bring to the surface of relevant historical data on accidents and propose preventive actions has made the system of an advantage to the training of new employees.

The most impressive perhaps is that Hacarus has collaborated with Tokyo Electron to introduce AI-enabled safety surveillance alerts to the semiconductor clean rooms- where any procedural deviation can bring up millions of dollars in wasted products. Their gaggle tracker software enables them to detect possible safety non-compliances much earlier and keep these sensitive facilities in line with strict hygienic safety standards.

The Small Data Advantage Goes Global

Hacarus has a great track record, support of the international community in recognizing its technology. In 2023, it introduced the first sparse modeling-based automobile inspection line in Europe, to a big German OEM which has cut inspection errors by 82% and computing resources needs by almost 75%. In early 2024, the company raised $28 million Series C financing with Mitsubishi Electric and industrial investors to leverage the energy sector, specifically on predictive maintenance. Moreover, Hacarus is also working to transform its businesses into medical diagnostics, where it has had initial success in detecting rare neurological disorders at Kyoto University Hospital.

Preserving the Future by Digitizing the Past

Hacarus resolves the so-called silver tsunami in the manufacturing sector of Asia capturing and digitizing the tacit knowledge of skilled workers who retire. This technology is especially worthy to Japanese manufacturers short of labor. As an example, Hacarus teamed up with a 300-year-old metalworking manufacturer and jointly produced an AI system that simulated the judgment of master craftsmen to ensure high standards were met but also improved the consistency of production. Hacarus is not another AI vendor, they provide a differentiated view on industrial automation that augments human knowledge, operate on constraints, and show that the absence of data can produce more knowledge.

Takashi Someda, CEO of Hacarus