Slingshot AI Accelerated Mental Health AI with Fine-tuning
A Together AI customer story describes how Slingshot AI trains Ash, reporting that its purpose-built therapy models outperform closed-source alternatives that cost five times as much, while serving more than 50,000 people.
In a customer story with Together AI, Slingshot AI explains how it trains the models behind Ash using a multi-stage pipeline of supervised fine-tuning and preference optimization on clinical conversations. The piece reports that Slingshot's purpose-built models deliver superior results to closed-source alternatives that cost roughly five times as much, while the team ships new models three to seven times a week, far faster than its earlier weekly cadence. It notes more than 50,000 people receiving support through Ash. Co-founder Daniel Cahn frames the work around speed and reliability, describing "the technical challenge of running our multi-stage pipeline reliably at the conversation lengths our therapy models require," while co-founder Neil Parikh stresses building AI that genuinely helps people. The collaboration aligns Slingshot with a well-known AI training platform, reinforcing the engineering depth behind the product.
