Behind Every Chip there is Algorithmic Brilliance
Advanced algorithmics, machine learning, and “crazy” ideas: Nova believes in the strategic importance of software
An engineering failure in a building, whether at the foundations or in a brick in one of its floors, may lead to the collapse of the whole structure. The same is true for the field of computer chips, except that here the physical dimensions are less than one nanometer – a billionth of a meter. When the cost of every manufacturing error is the loss of billions of dollars, even the slightest component irregularity may disable the entire chip, cause tremendous losses, and disrupt a huge variety of products and services. At Nova, an Israeli company that offers leading inline metrology solutions, advanced hardware and sophisticated software are combined to solve this issue.
Leading semiconductor manufacturers use Nova’s measurement and monitoring systems, to minimize expensive mistakes in the chip manufacturing process. To support the growing demand for higher computation power and larger memory capacities, the Semiconductor development process is becoming more advanced and complex, requiring the measurement techniques to become increasingly more sophisticated.
“We support companies in the development and manufacturing process, conduct experiments, and use the data provided by the customer to form models,” says Dr. Danny Kandel, VP of Technology in the CTO group. “We are often required to develop unique and innovative metrology solutions, to support the development of new advanced integrated circuits. Since we deal with critically small dimension ranges which are difficult to measure, the measurement techniques must combine hardware, advanced algorithms, and machine learning.”
Dr. Danny Kandel, VP of Technology
Kandel is part of Nova’s CTO division, dealing with future technologies and focuses on the effort to solve future challenges in a rapid manner. Dr. Dror Shafir, VP of New Technology, in the CTO division: “for SSD memory chips, for example, we had to develop an optical technology combined with unique algorithms, able to measure complex signals, filter unnecessary data, and noise, and analyze the input aiming to provide an accurate representation to our customers. New innovative algorithmics, combining the conversion of scientific theories with complex mathematical models, were formed in the development process.
How can machine learning help achieve the best results with unique hardware?
The greatest ideas say Kandel and Shafir, arise when experts in different fields in the company brainstorm together. “Our clients require quick solutions for complex problems, and our way of overcoming these challenges is to get physicists, algorithm developers, and software experts to think together.” This, for example, is the reason why Nova started using machine learning as an alternative to traditional physics computations; “the machine learning applications we have embedded improved hardware performance, dramatically shortened the time required for highly-complex problem solution, and simply took over the market.”
The next step would be combining physics-based and machine-learning-based algorithms. “We have in fact created a new type of algorithm with distinct advantages of accuracy and speed, allowing us to solve problems which used to be unsolvable until recently.” They both affirm that this innovative and creativity-encouraging approach, is deeply rooted in the company’s DNA: “whenever a team member comes up with a good idea, no matter how crazy or expensive it seems to be, everyone – from the CEO to junior directors – tells them to go for it, and it happened more than once. The potential impact of it is huge.”
Software Takes the Front Seat
Nova’s software group in the Dimensional Metrology division works closely with the algorithm, hardware, and system groups, and uses a wide variety of technologies to support the company’s growing product demand. “On the one hand, we offer traditional technologies that are at the core of Nova’s offering, technologies that we are constantly developing. On the other hand, we are adding top-tech software modules to keep up with the rapid development in our industry,” explains Kenny Krupnik, VP of Software in the Critical Dimensions Division, “We work with a wide variety of technologies and each of our development teams can offer a supplementary solution.”
Nova offers solutions of various types: software-hardware combined products dealing with the physical aspects of the metrology process as well as with robotics and communication, big data management products, tool-fleets management, software, and machine learning. “Historically, the chip market has been hardware-oriented, while Nova has an exceptional strategic approach of leading the market through software,” Krupnik adds. “For example, one of our flagship products, Nova Fit, is a leading machine-learning product that plays a major role in allowing Nova to meet its clients’ most strict measurement requirements. Our goal, producing advanced software solutions which will hasten the company’s growth, has proven itself during recent years, as the profit from these products showed significant growth.”.
Dr. Dror Shafir, VP of New Technology
“The chip industry stands is at the core of every other industry,” Shafir says. He states that the more complex the products get, the more advanced the chip and its manufacturing monitoring solutions must get: “we have to reinvent ourselves every time because we can’t just exploit more of the same thing, it just wouldn’t be enough.” Kandel adds: “if you look for challenges and openness to modern innovations, especially in the algorithmics field, with an open and supportive work environment, you will find it here.” And Krupnik sums up: “unlike most of our competitors, Nova’s size is optimal, producing a great combination of a strong, steadily growing company, with a progressive, family-like organizational culture, which allows personal growth and a never-ending technological experience. Things move fast and it’s a fun environment to work in.”
