Press · podcasts · speaking · collaboration
I speak about designing private AI around organizations, deterministic boundaries for consequential agent actions, customer-controlled infrastructure, and the evidence organizations need before a promising demo becomes a dependable system.
Verified biography
Aki is the professional name; books and research papers use Akhilesh Gogikar. Both names refer to the same person.
Short bio
Aki Gogikar is the founder and Executive Vice President of Mendel Info Labs. He leads product and engineering across OneNew and ActPass, building private, controlled AI services and separate, deterministic governance for consequential agent actions.
Research bio
Akhilesh (Aki) Gogikar builds auditable AI systems and studies machine learning optimization. His public work includes a compute-matched study of adaptive Runge-Kutta methods for Adam, published on arXiv with code, data, and figures.
Media asset
A square, high-resolution portrait is available for event pages, articles, and podcast artwork. Credit: Aki Gogikar.
Download portraitConversation starters
These topics work for technical, executive, and regulated-industry audiences. I favor concrete tradeoffs over generic AI predictions.
Private AI services
Why privacy-aware teams need tailored workspaces, approved data and tools, and ongoing service — not another generic model subscription.
Agent security
Why consequential actions need deterministic permission checks before they execute.
Private deployment
How customer-controlled infrastructure changes deployment, operations, and accountability.
Research practice
What compute-matched experiments reveal when a sophisticated optimizer does not generalize better.
Public work
Start with the products, paper, and reproducibility trail rather than a list of unsupported superlatives.
Products
OneNew's private-AI platform direction and ActPass's independent permission boundary for consequential agent actions.
Explore the productsPaper
A compute-matched study of adaptive Runge-Kutta step control in machine learning optimization.
Read on arXivCode & data
The public code, data, figures, and research artifacts supporting the paper.
Open GitHubPress · podcasts · panels · technical reviews
Share the format, topic, timing, and what you want people to understand when the conversation ends. I read every note personally.
Email aki@onenew.ai
