You can now train your own Decision model like Jev locally! 💡
You can now train your own Decision model like Jev locally! 💡 We increased Qwen3.5 0.8B’s accuracy from 20.7% to 74.3% across 3 decision benchmarks - on just 4GB VRAM. Turn any LLM like Qwen3.8, Gemma 4 into decision models with our open-source repo.
Unsloth AI GitHub: https://lnkd.in/gyaDBTxK Guide and Notebooks: https://lnkd.in/gX-_5mJ8
We fine-tuned with a Clef head using Unsloth and LoRA (r=64) for one epoch, increasing downstream accuracy from 30–37% to 78%.
Why this matters for talent
Helm.ai publishing open source tooling that lets developers train their own decision models locally is a genuine technical signal, showing the company's research team is active and willing to share work publicly. Helm.ai does not currently show tracked open roles on the board, so this is best read as evidence of engineering culture and technical depth for candidates researching the company ahead of applying directly.
5,568 likes · 189 comments · 478 shares · Read the original LinkedIn post. Curated by Larry Sherwood Jr. from company announcements; analysis above is my own.