private AI · controlled deployment · agent governance
I'm Aki Gogikar, founder and EVP of Mendel Info Labs and builder of OneNew and ActPass. I design controlled AI systems for privacy-aware organizations — connecting approved data, private workspaces, appropriate tools and models, and human oversight.
What I build
Privacy-aware organizations need more than a public-model subscription. They need a private environment designed around approved data, workflows, policies, and people — delivered as one coherent customer experience.
Private AI services · onenew.ai
OneNew is the single customer-facing offer for private AI built around each organization. Its service model is designed to discover the requirement, design the controlled solution, connect approved data, provision a managed private-AI Workspace on customer-controlled infrastructure, and operate it with the customer. Its broader direction spans Workspace, API, Desktop, organizational memory, private model access, and approved custom tools.
Visit OneNewIndependent governance · actpass.org
ActPass is a separate, vendor-neutral governance product for AI agents. It evaluates consequential actions as allow, deny, or needs approval before execution.
Visit ActPassCompany · mendelinfolabs.com
Mendel Info Labs builds and operates trust-first AI products, including OneNew and ActPass. I'm its founder, and I lead product and engineering across both.
Visit Mendel Info LabsResearch & writing
I publish methods, code, and results so the work can be challenged and reused. My interests span auditable AI, agent boundaries, private infrastructure, optimization, and complex systems.
Research paper · arXiv · 2026
A compute-matched machine learning study asking whether adaptive Runge-Kutta optimization improves generalization, with the method, limitations, and mixed results made explicit.
Read the paperOpen research · GitHub
Code, data, figures, and the reproducibility trail behind the RK-Adam paper. The repository makes the compute accounting and negative results inspectable rather than hiding them.
Explore the codeEssays · Medium
Notes on artificial intelligence, data, systems, and the practical work of turning technical ideas into decisions people can use.
Browse the archiveBook
How simple rules produce complex behavior — in systems, markets, and machines. Published under my author byline, Akhilesh Gogikar.
Looking for an interview guest, speaker, research collaborator, or technical reviewer? Start with the press and speaking page.
About
I'm Akhilesh (Aki) Gogikar — founder and EVP of Mendel Info Labs.
I go by Aki professionally. My books and papers carry my full name, Akhilesh Gogikar — same person, and this page exists partly so you don't have to wonder.
I lead product and engineering across OneNew and ActPass. My work asks how to give privacy-aware organizations tailored AI capability while keeping their data, consequential actions, and decision evidence under meaningful human control.
I write about private AI systems, agent security, organizational memory, efficient inference, and dynamical systems. I aim to make difficult technical and governance choices easier to discuss, test, and defend. If you're working on any of that, I'm easy to reach and happy to compare notes.
private AI · controlled deployment · governance · collaboration
Working through a difficult private-AI, deployment, or agent-governance problem? Tell me what you're building and what you're accountable for. I read every note, and if OneNew or ActPass isn't the right answer, I'll say so.
Email aki@onenew.ai