Building a research team is not a hiring exercise. It is an architectural decision — one that shapes what problems you can solve, how fast you can solve them, and whether your solutions survive contact with reality.
The problems we're solving
Neuraphic operates at the intersection of three research domains that rarely overlap: adversarial machine learning, real-time systems engineering, and cloud infrastructure security. Most organizations specialize in one. We need people who can work across all three — or who go so deep in one that they redefine what's possible within it.
Our adversarial AI research focuses on understanding how language models can be manipulated at inference time, and building systems that detect and neutralize those attacks in under two milliseconds. Our infrastructure security research asks whether AI can reason about vulnerabilities the way a human researcher does — not through pattern matching, but through contextual understanding of how systems fail.
These are not incremental problems. They require people who are comfortable with ambiguity, rigorous under uncertainty, and willing to publish negative results.
What we value in researchers
We do not hire for credentials. We hire for judgment. The ability to identify which problems matter — and which are distractions — is more valuable than any publication record. We look for people who have built systems that work in production, not just systems that work in papers.
We value clarity of thought over volume of output. One well-designed experiment that reveals something true is worth more than ten papers that confirm what everyone already suspected. We want researchers who can explain their work to an engineer and have that engineer build something real from it by the end of the week.
We also value intellectual honesty. If a method doesn't work, we need to know early. If a result is ambiguous, we need to say so. The systems we build protect other people's infrastructure — there is no room for optimistic interpretation of marginal results.
What we value in engineers
Our engineering challenges are unusual. We build real-time classification systems that must process every request to an AI model without adding perceptible latency. We build distributed infrastructure that must remain secure even when individual components are compromised. We build zero-trust systems where no service trusts any other service by default.
This requires engineers who think about failure modes before they think about features. Who understand that the difference between 1.4 milliseconds and 14 milliseconds is the difference between a viable product and an unusable one. Who can write code that will be audited by people trying to break it.
How to reach us
We are currently hiring across research, engineering, and security. Our process is direct: we share the problem we're working on, you show us how you'd approach it, and we decide together whether the fit is right. No whiteboard puzzles. No take-home assignments designed to waste your time.
If this sounds like work you want to do, we'd like to hear from you.