While companies like OpenAI spend billions building massive AI infrastructure, two brothers from Sydney are proving that smart engineering can sometimes beat brute force.
Meet Unsloth AI founders Daniel Han and Michael Han — the Australian duo building one of the most important open-source AI tools in the world right now.
From NVIDIA Engineer to Open-Source AI Innovator
Before starting Unsloth, Daniel Han worked at NVIDIA, where he focused on optimizing machine learning algorithms.
One of his standout achievements was accelerating the t-SNE algorithm by nearly 2000x — a massive improvement for data visualization and machine learning workflows. He also reduced SVD memory usage dramatically, making heavy computations more efficient on limited hardware.
But what really made people notice him was something even bigger:
He discovered and fixed more than 20 bugs inside some of the world’s most important open-source AI models, including:
Llama
Gemma
Mistral
Phi
These were bugs that large AI labs themselves had missed.
The Birth of Unsloth AI
In December 2023, Daniel and his brother Michael launched Unsloth AI with a simple but ambitious goal:
Make large language model (LLM) training dramatically faster and cheaper for everyone.
Instead of relying on billion-dollar GPU clusters, Unsloth focused on software optimization, custom kernels, memory-efficient algorithms, and low-level GPU engineering.
The results shocked the AI community.
According to the company and Y Combinator profiles, Unsloth enables:
LLM fine-tuning up to 30x faster
70–90% lower memory usage
Training on consumer-grade GPUs instead of expensive enterprise hardware
Faster reinforcement learning and inference workflows
Open Source First — And Completely Free
Unlike many AI startups chasing enterprise lock-ins, the Han brothers released Unsloth as fully open source.
That decision helped the project explode in popularity across the developer community.
Today, Unsloth has:
Over 64,000 GitHub stars
Around 10 million model downloads every month
Adoption by organizations including NASA and Canva
Backing from Y Combinator as part of the Summer 2024 batch
A tiny team of roughly 8 people shipping cutting-edge AI infrastructure
And perhaps the most surprising part:
They achieved all of this after raising only about $500,000 in seed funding.
Why Unsloth Matters
The modern AI race is often framed as a competition of money and hardware.
Bigger data centers. Bigger GPU clusters. Bigger funding rounds.
But Unsloth represents another path:
better engineering.
By optimizing training efficiency instead of simply scaling hardware costs, the Han brothers are helping democratize AI development for students, startups, researchers, and independent developers worldwide.
Their work lowers the barrier to entry for AI innovation — something increasingly important as training costs continue to rise.
Small Team, Massive Impact
The story of Unsloth is becoming one of the most inspiring examples in open-source AI.
Two brothers from Australia. A tiny engineering team. Minimal funding. Global impact.
While tech giants spend billions building AI infrastructure, Daniel and Michael Han are proving that intelligent optimization and open-source collaboration can still reshape the industry.
Absolute legends.