Applied Compute

Applied Compute builds what it calls Specific Intelligence: custom models trained on the latent knowledge inside individual enterprises, tailored to each client's own needs and competitive advantages. The company reports state-of-the-art performance on customer evaluations. Alongside the models, it deploys proprietary in-house agent workforces within client organisations - validating and deploying them in days rather than months - with deployments live at Fortune 50 companies. Named customers include DoorDash, Cognition and Mercor, spanning food delivery and logistics, AI software, and talent and recruiting.

The technical work covers reinforcement learning, post-training, ML systems infrastructure for RL training, agentic AI, custom model training, enterprise deployment, agent platforms and tooling, and evaluation and model validation. Training stacks, agent platforms and tools are built entirely in-house by Applied Compute engineers, which the company describes as a means of enabling rapid iteration and continuous improvement. Engineers embed directly within client teams rather than outsourcing or delegating the work.

Applied Compute has raised $80m in total funding, with backing from Benchmark, Sequoia and Lux Capital. Two-thirds of the team are former founders, and the roster includes International Math Olympiad winners and top AI researchers; founding-team experience includes work at OpenAI on Codex, the o1 reasoning model and RL training infrastructure.

For marketing engineers, the company's public footprint rests on enterprise-grade custom model and agent deployments, an in-house-built toolchain, and a customer base of large technology and consumer platforms.