I occasionally work with teams applying the Living Framework research to real AI workflows. This is not a consultancy. Engagements are selective and research-grounded.
This is not a consultancy. There is no product, no retainer, no packaged service. I do not sell prompt engineering templates or AI implementation playbooks. What I offer is the application of two years of documented research to a specific problem in a specific context — directly, without intermediaries.
I work with teams where the reliability of AI collaboration has real consequences — where errors propagate into real decisions, where context loss causes real cost, where the absence of verification has already caused problems or is likely to.
The research is most useful where AI is being used in sustained, complex, high-stakes work: long-horizon projects, knowledge-intensive workflows, situations where the human and the AI are collaborating over weeks or months rather than hours. If AI is being used as a search engine or spell-checker, the framework is probably overkill.
I am also interested in conversations with researchers working on adjacent problems — reliability, verification, human-AI interaction, extended cognition. If the papers are relevant to work you are doing, a conversation is worth having.
Microsoft Azure’s production AI platform now runs a four-layer closed-loop verification architecture — a system for catching and correcting errors before they propagate into decisions, achieving above 90% autonomous resolution at scale. The mechanism is architectural, not a capability question.
That is the same problem this research addresses from the other direction: what structural conditions need to be in place for AI collaboration to remain reliable over sustained, high-stakes work? The failure modes documented across two years of case studies — context drift, undetected degradation, verification gaps — are the same problems that deployment at scale eventually forces you to solve.
Teams rarely discover the reliability problem in the abstract. They find it when a workflow fails, when an error compounds undetected through a long-horizon task, when the output looks confident but isn’t grounded. The research exists so teams can build the architectural response before that discovery is expensive.
Email is the right channel. Share a bit about what you are working on and how the research connects to it. I am genuinely interested in these conversations and respond to every message.
Describe your context, your problem, and why the research seems relevant. That is enough to start a conversation.
rishisood@protonmail.comIf you are not ready for a direct engagement but want to apply these principles to your own work, I have authored twelve field-specific AI guides — written for managers, lawyers, teachers, HR professionals, recruiters, and administrators. Each guide is built around prompts grounded in the same research findings, designed for reliable, structured AI collaboration in your specific context.