San Francisco, CA – Yash Patil, CEO and co-founder of Applied Compute, recently articulated the core thesis behind his company's rapid ascent, emphasizing the strategic imperative for enterprises to transform internal workflows and domain knowledge into proprietary AI systems. This vision has seen Applied Compute achieve a valuation of $1.3 billion within a year of its founding. Patil's perspective underscores a shift towards specialized AI that leverages unique organizational data for competitive advantage.
"Companies need to turn their workflows, domain knowledge, and accumulated judgment into AI systems that improve with each use," Patil stated in the tweet. He further elaborated on the necessity of "private evals" to gauge model improvement against business outcomes, rather than relying solely on external benchmarks. This approach allows AI models to grow stronger on real data traces from within the organization.
Applied Compute, founded in early 2025 by former OpenAI researchers Patil, Rhythm Garg, and Linden Li, specializes in "Specific Intelligence" for enterprises. The company develops custom AI agents trained on proprietary data, integrating directly with internal teams to bypass reliance on general-purpose AI. This focus aligns with Patil's belief that a firm's knowledge base makes institutional memory queryable and token usage more efficient.
The startup has experienced significant financial growth, securing $20 million in a seed round at a $100 million valuation in June 2025, followed by an $80 million raise that pushed its valuation to $700 million by October 2025, and subsequently to $1.3 billion by May 2026. This rapid appreciation reflects investor confidence in its reinforcement learning-based approach. Key customers include DoorDash, Cognition AI, and Mercor, utilizing Applied Compute's technology to build specialized systems.
Patil views this continuous improvement cycle as the "new IP of the firm," describing it as a "hill climbing machine" that compounds over time. He explained, > "Every improved workflow generates better training signal, which accelerates the accumulation of tacit knowledge unique to the firm." This proprietary feedback loop creates an advantage that is difficult for competitors to replicate, irrespective of advancements in general AI models.
