The 24-year-old founder and CEO Carina Hong founded Axiom Math in March 2025 and recruited a team of ten employees, most of whom come from Meta to construct a Math AI model.
Axiom Math
LAut autumn, Carina HongA doctoral student at Stanford University spent her weekends with mathematical research on a Verve Coffee Roasters near the campus. At certainly one of the common tables of the café, she hit a conversation with Shubho Sengupta, a AI researcher at Meta, who worked on large voice models that wrote software tests. They talked in regards to the interface of their respective fields for hours and the opportunity of developing a AI that would solve the hardest mathematical problems on the earth and discover recent ones.
Shortly afterwards, Hong began from Stanford to start out Axiom Math, a startup of the early stage that goals to accumulate a so-called “AI mathematician”-a model that may solve complex mathematical inquiries to generate detailed evidence of the several steps to get the reply and check his own work. The idea is to convert English -language mathematics from textbooks, archived work and magazines right into a software program in order that the AI ​​can generate recent problems and the solutions may be formally tested and verified. She hopes that the model will ultimately create completely recent knowledge by suggesting assumptions, theories which are too true but haven’t yet been proven mathematically.
“Mathematics is the perfect sandpit for building superintelligence,” said Hong to Forbes. Your startup value 300 million US dollars has collected seed promotion of 64 million US dollars with B capital with participation of risk corporations reminiscent of Greycroft, Madrona and Menlo Ventures.
In lower than a yr, Hong has arrange quite a lot of experienced tech veterans for his or her young startup – many from Metas fundamental ai research (fair) laboratory. The ten full -time employees of the startup include Francois Charton, who previously examined the usage of large -scaling models for mathematics and theoretical physics at Meta and solved a 100 -year -old math problem last yr. Aram Markosyan, a AI research scientist who led to META security and fairness research; And Hugh Leather, a former Metaai scientist who was certainly one of the primary in the sphere, used Deep Learning for the codegen. The movements come after META triggered 100 million US dollar Pay packages to poach the highest KI talent of Openai and other border -ai -Ai -Ai Labor for its superintelligence unit. But also in the course of the highest -class setting stroll, the social media giant has continued to lose essential AI researchers. Forbes reported.
For many researchers, the mission of Axiom was to make use of artificial intelligence for the mathematical discovery – within the office of the startup Palo Alto, where conference rooms are named after phenomenon reminiscent of Carl Friedrich Gauß and Ada Lovelace – a very important sales argument. “For me it was the opportunity to make AI for mathematics in a company that was serious and did not consider it a side quest,” said Charton.
Hong himself has a formidable background, filled with awards. The 24-year-old founder grew up in Guangzhou, China, and loved mathematics. She made her bachelor’s degree in mathematics and physics from where she wrote nine research and took 20 courses on advanced math. In 2023 she won the coveted Frank and Brenie Morgan Prize – the very best award that was granted to a student for his or her outstanding research on number theory and probability. As Rhodes scholars at Oxford University, she received a master’s degree in computational neurosciences.
Despite her progress with Axiom, Hong has steep competition. Ai Goliaths reminiscent of Openaai and Google Deepmind recently achieved gold medal level on the international mathematics Olympics, one of the crucial renowned mathematics competitions after their AI models had solved five out of six extremely difficult questions. But Hong said these benchmarks may be written off and don’t reflect math at research level.
At the moment, Axiom is specializing in training models that find recent mathematical problems and may solve them accurately. However, the researchers hope that their work can finally be applied to other areas, reminiscent of: “The solution to complex mathematical problems was the central part of so many human inventions,” said capital partner Yan-David Erlich. “It is crucial to create new problems that model reality and then solve it for the further development of human knowledge.”
More from Forbes
