The studio, and where it comes from.
Hermes Labs is an AI reliability engineering studio focused on retrieval, memory, agents, auditability, and the language layers around AI systems. We find the structural failures standard evaluations miss, where a model stays fluent and plausible while quietly distorting the evidence, the instructions, or the memory underneath, and we engineer them out. We call the practice Epistemic Engineering.
Hermes Labs was founded by Rolando Bosch.
Rolando, who goes by Roli, came to AI from the philosophy of language, not from machine learning. He studied philosophy, beginning with Wittgenstein and the question of how meaning actually works, then moving through phenomenology and hermeneutics toward a single problem that still drives the studio: what do “understanding” and “truth” mean for a system that only ever sees language?
That question is where the name comes from. Hermes Labs is named for hermeneutics, the philosophy of interpretation and understanding, and beyond it for Hermes, the Greek god of messages and translation. The throughline traces to Hans-Georg Gadamer’s Philosophical Hermeneutics. The whole book rewards the effort, but David E. Linge’s introduction alone is one of the clearest things written on what understanding actually is, in people or in machines.
The work began at LPCI Innovations, exploring language as interface and as infrastructure: early demos like LucidiGPT and AI Ching, and consulting for teams who needed AI to be legible rather than magical. A parallel series of videos on AI and philosophy sharpened the thesis, and that thesis became Hermes Labs: a studio that treats the language layer as where AI systems are made reliable, or left to fail silently.
Before AI, Rolando worked in political organizing and community advocacy. The lesson that carried over was how to move a dynamic system you do not fully control: when the point of leverage you need is closed to you, and the one you can reach behaves like a black box. An LLM interaction is a dynamic system of exactly that kind.
Born in Galicia, Spain, raised in Miami, and now based in San Francisco, Rolando grew up moving between places and languages, which makes one thing impossible not to notice: how much of understanding depends on context, and how easily two people working from different contexts come away from the same words with different meanings. Once you see it in people, it is everywhere in AI systems too.