She Raised $64M to Build an AI Math Prodigy | Carina Hong, CEO of Axiom

2026-02-05 · Source: Weights & Biases · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences, Software Development & Engineering · Depth: Advanced, extended

Summary

Axiom Math, led by CEO Karina Hung, has raised $64 million to develop a self-improving AI reasoning engine that combines generation and verification, starting with an AI mathematician. The system utilizes formal languages like Lean to ground natural language, enabling higher sample efficiency. Axiom's architecture includes a prover, a conjecturer, and a knowledge base, all woven together by an auto-formalization system. This AI recently achieved a score of 8 out of 12 (later 9 out of 12) on the 2026 Putnam Exam, a notoriously difficult mathematics competition for undergraduates, surpassing the top human score of 9 out of 12 from the previous year. The company aims to apply this technology beyond mathematics to areas like hardware and software verification, code migration, and database consistency, addressing the critical need for provable guarantees in safety-critical domains.

Key takeaway

For CTOs and engineering leaders evaluating AI for critical systems, Axiom's approach demonstrates that combining probabilistic AI with deterministic formal verification can achieve unprecedented reliability and performance in complex problem-solving. You should consider how such hybrid AI systems could provide provable guarantees for your safety-critical applications, reducing manual verification burdens and accelerating development in areas like chip design, secure coding, and robust database management.

Key insights

Axiom Math's AI combines formal verification with probabilistic generation to solve complex math problems and aims for broader application.

Principles

• Combine generation and verification for robust AI reasoning.
• Formal languages enhance AI sample efficiency and provable guarantees.
• AI can act as a "diligent grad student" for human intuition.

Method

Axiom's system integrates a prover, conjecturer, and knowledge base, using auto-formalization to translate natural language into formal Lean code, enabling rigorous, verifiable mathematical proofs and discoveries.

In practice

• Apply formal verification to hardware/software design.
• Use AI for code migration and legacy system equivalence checks.
• Enhance database consistency with formal proofs.

Topics

• AI Mathematician
• Formal Verification
• Reasoning Engines
• Lean Language
• Putnam Exam

Best for: Research Scientist, Investor, CTO, AI Scientist, AI Engineer, Entrepreneur

Related on AIssential

• 🔬Scaling Past Informal AI - Carina Hong, Axiom Math — Verified AI, grounded in formal mathematics, scales human brilliance and offers significant performance advantages over informal methods.
• 😸 AI just solved unsolvable math — AI can now autonomously discover new mathematical truths, moving beyond merely solving known problems.
• She Raised $64M to Build an AI Math Prodigy | Carina Hong, CEO of Axiom — Axim combines formal languages and AI to create a self-improving reasoning engine for mathematical verification.
• Inference: When Will We Solve Intelligence — Solving intelligence requires a hybrid approach combining formal verification with large language models for robust reasoning.
• Google tops OpenAI's math breakthrough — 9 to 1 — AI systems can now autonomously generate and formally verify solutions to complex, long-unsolved mathematical problems.
• Building an AI Mathematician [Carina Hong] - 754 — AI mathematicians combine AI, programming languages, and math to formalize and verify proofs, addressing data scarcity via autoformalization and synthetic data.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Weights & Biases.