MuukTest announced a new AI agent designed to automate much of the software QA testing process on Tuesday. The Raleigh, North Carolina-based startup aims to reduce test creation to a click. “Since we started, our vision has been automating software QA technology to reduce test creation to a click,” said co-founder and CEO Ivan Barajas Vargas.
These tools enable testers to check every menu, button, and operation in the software user interface under multiple conditions. The goal is to catch as many bugs as possible before release. CTO and co-founder Renan Ugalde stated that Barajas Vargas has over 20 years of experience in software quality assurance testing.
They have used Ugalde’s engineering expertise to create an AI agent that assists in building test suites. This AI leverages multiple large language models, traditional machine learning, computer vision, and image recognition to mimic the decision-making process of a QA tester. “We trained AI agents to think just like a QA tester, to understand the context within the application — knowing what a menu is, what an input is, and when you expect to see something,” Ugalde explained.
This involves reinforcement learning and the founders’ extensive QA experience to develop the AI agent. For MuukTest, this AI acts as an intelligent assistant, performing many of the mundane tasks traditionally handled by human QA testers. The company’s goal has always been to reduce the effort required to generate and run QA tests.
Early versions of their solution utilized no-code algorithms to create the tests. With the new generative AI product, customers can simply describe the type of test suite they need, and MuukTest creates it automatically. MuukTest, founded in 2019, began to find product-market fit early last year.
Even before adding AI elements, the company experienced a 15x revenue increase over the prior year. They believe their new capabilities will enable even faster growth. The company, which launched through a Massachusetts-based startup incubator, has raised a total of $6 million in investments and grants.
With 36 employees and 10 contractors, Barajas Vargas emphasized that the company intends to remain conservative with spending. The new AI agent capability is generally available starting today. Generative AI has the potential to significantly impact software development and productivity, but with this increased productivity comes the amplified challenge of software testing.
If five or ten times the amount of code can be generated as before, that means significantly more code needs to be tested. “Many CFOs are currently eyeing investments like GitHub Copilot or similar products, considering a cost of $30 per month per developer,” said Jim Scheibmeir, a senior director analyst.
MuukTest debuts automated QA tool
“However, we often overlook that a bottleneck in software development isn’t just writing code but also testing it. We’re going to make developers much more productive, which could include an increase in writing defects.”
Unlike AI-assisted development tools that make developers generate more code, AI-assisted testing tools aim to reduce the necessity for extensive testing. For example, test impact analysis tools can create a testing strategy properly sized for the actual code change being pushed, running only essential tests rather than every test available.
“These tools provide focus for testers,” Scheibmeir explained. “It’s difficult to give testers focus today, as there’s often a pervasive feeling necessitating the testing of everything, even under time constraints.”
Arthur Hicken, a chief evangelist, agrees that test suites now frequently take hours or even days to complete. Generative AI’s role in optimizing test coverage can alleviate this problem.
“AI can provide quite a good estimation of what needs validation for a change,” Hicken said. Generative AI is also integrating into various aspects of the software testing process, from test generation to management, making the testing lifecycle more efficient. As generative AI continues to make strides in software development, it equally demands innovation in testing methodologies.
The balance between increased productivity and efficient testing will be crucial for future advancements. In the rapidly evolving landscape of semiconductor engineering, continuous advancements in test, measurement, and analytics are essential. As technology progresses towards more advanced nodes and packaging, industry experts are addressing both imminent and long-term challenges.
Recent developments have primarily focused on gallium nitride power devices, with best practices still emerging. However, significant strides have been made in turning reactive root cause analytics into proactive monitoring systems. Prasad Bachiraju from Onto Innovation emphasizes the importance of such systems in preventing the propagation of defective die.
Faisal Goriawalla of Synopsys has delved into the High Bandwidth Memory (HBM4) test challenges. His insights underline the necessity of detecting defects before system failures occur, ensuring the reliability of the semiconductor devices. In the automotive sector, the push towards smaller process nodes is a significant trend.
Fisher Zhang from Teradyne highlights the auto industry’s need for early and continuous testing engagement due to the increasing complexity of these nodes. Artificial intelligence (AI) and machine learning (ML) are now pivotal in enhancing testing processes. Ira Leventhal of Advantest demonstrates how AI and ML algorithms can identify patterns and anomalies that might elude human testers or traditional methods, thereby improving test accuracy and efficiency.
These insights collectively demonstrate the industry’s ongoing efforts to enhance semiconductor testing and measurement. As technologies and methodologies continue to develop, the importance of proactive strategies and advanced analytical tools becomes increasingly evident.
