An independent research initiative focused on reliable, verifiable AI systems.
As AI agents become autonomous — reasoning, acting, coordinating — the question shifts from "what can AI do?" to "can we trust what it does?" We publish open research and build open-source tools for AI trust, reliability, and verification. Our work stands on its own merits — reproducible methods, public code, and rigorous evaluation.
Founded
Open-Source Packages
Divisions
Open Source
We publish on AI reliability, evaluation methods, and verification — sandbagging detection, adversarial evaluation, reasoning verification, and trust dynamics in multi-agent systems.
We build and maintain libraries for trust-aware AI systems. All packages on PyPI and npm, AGPL-3.0 licensed.
Packages
Research Areas
Languages
Context intelligence for AI agents
Sandbagging detection via activation probes
Steering vectors for runtime behavior control
Neuro-symbolic verified code synthesis
Field-theoretic memory for AI agents
Evolutionary adversarial testing
Our research is guided by a few core principles that shape everything we build and publish.
Every claim is backed by code. Our open-source packages implement the methods described in our papers. If you can't reproduce it, it doesn't count.
We assume AI systems will be deployed in adversarial conditions — gamed benchmarks, strategic underperformance, coordinated deception. Our methods are designed to hold up under these assumptions.
LLM evaluation is plagued by noisy benchmarks and overfit leaderboards. We develop evaluation methods grounded in statistical testing, distributional analysis, and formal verification where possible.
Research is published openly. Code is open-source. We believe trust research must itself be trustworthy — and that starts with transparency.
Rotalabs was founded by engineers and researchers with backgrounds in large-scale distributed systems, cloud infrastructure, and AI/ML — with prior experience at Google and in production ML systems serving hundreds of millions of users.
Our team combines deep systems engineering expertise with applied AI research. We've built and operated infrastructure at hyperscale, and we bring that same rigor to the problem of AI reliability and trust.
We are currently in stealth and will share more about the team as we move into our next phase. In the meantime, our research and open-source work speak for themselves.
Rota, Inc. is a Delaware C-Corporation headquartered in San Francisco. Research and products are organized across three divisions.
Our research is open. Applied work builds on published methods.
We're looking for people excited about AI reliability, evaluation, and systems engineering. If you want to work on hard problems in the open, we'd like to hear from you.
Interpretability, adversarial ML, formal verification, multi-agent trust. Publish openly and build tools from your research.
ApplyBuild trust infrastructure at scale. Python, Rust, distributed systems, GPU kernels. Production ML experience preferred.
ApplyAll our packages are open source. Issues, PRs, and discussions welcome across 12 repositories on GitHub.
GitHubType to search across all pages and posts
Press ↑ ↓ to navigate, Enter to select