Personal and distributed intelligence
How intelligence can improve for an individual without that individual's raw context being consolidated into a single platform.
RESEARCH & PARTNERS
Researching the computing models personal AI will need next.
Prifina is building the personal data and intelligence layer for AI's next era. The parts that are not yet solved are being worked on in the open, with research partners.
What is different: the research questions are stated as questions, not as shipped features. What to do next: bring a question, a capability or an infrastructure position.
RESEARCH COLLABORATION
Exploring the computing models personal AI will need next.
University of Oulu, Future Computing Group. This is Prifina's formal research relationship, and it is separate from the wider academic community using our products.
Prifina collaborates with the University of Oulu's Future Computing Group to explore personal and distributed AI, edge intelligence and governed computing across devices, regional infrastructure and cloud environments.
WE ARE RESEARCHING: An open research direction. No availability implied.Future Computing Group (opens in a new tab)
The collaboration is presented as text because it is a working relationship, not a logo placement. No other organisation on this site is presented as an endorsement.
ACADEMIC COMMUNITY
Used by people working across research, education and professional knowledge.
Individual professors, researchers, educators and other professionals associated with universities around the world have created or used Prifina products, primarily AI Twin. Their use helps demonstrate how personal AI can make specialist knowledge more accessible without turning it into a generic, uncontrolled chatbot.
Institutional affiliations of individual users are shown with permission where required. The logos do not imply endorsement, procurement or a formal relationship with the institution.
INDUSTRY CONVERSATION
Taking personal AI into wider industry conversations.
Prifina and Mastercard introduced Personal AI Twins at Uplift 2025, presenting individually controlled AI to leaders across banking, fintech, retail and commerce.
A specific collaboration around one event. It is not a customer relationship, an adoption or an endorsement.
Uplift 2025
Personal AI Twins, introduced to leaders across banking, fintech, retail and commerce.
RESEARCH QUESTIONS
Five questions the next architecture has to answer.
- Q01
How can AI learn with individuals without pooling all raw personal data?
Centralised training made the first era possible. Deeper personal intelligence needs richer individual context. That is exactly the context least appropriate to pool. We are studying learning arrangements where the signal travels and the raw record does not.
- Q02
How should permissions for AI agents be expressed and enforced?
An agent acting for a person needs a scope narrow enough to be safe and legible enough to be reviewed. We are examining permission grammars that cover purpose, duration, data class, audience and revocation.
- Q03
How can provenance remain inspectable across several AI models?
When capabilities are interchangeable, provenance cannot belong to one provider. We are exploring how source attribution can be carried by the personal layer rather than by the model.
- Q04
Which tasks belong on devices, trusted clouds, regional infrastructure or larger cloud systems?
Latency, sensitivity, cost, jurisdiction and capability all pull differently. We are mapping which personal AI workloads are genuinely sensitive to placement and which are not.
- Q05
How can shared intelligence be useful without dissolving the personal boundary?
A family, a team or a neighbourhood benefits from shared context. We are studying contribution models where a person adds a bounded slice to a shared space and keeps the rest personal.
RESEARCH THEMES
Where the work is concentrated.
Edge and regional computing
Which parts of a personal AI workload belong on a device, in regional infrastructure, or in a larger cloud environment, and what decides that.
AI agent governance
How an agent's scope, permissions and expiry can be expressed in a way both a person and a system can verify.
Permissioned learning
How learning signals can be approved, bounded and withdrawn instead of being an invisible byproduct of use.
Provenance and inspectability
How a person can see which source produced an answer when several models and processing environments were involved.
Shared intelligence without pooling all raw personal data
How a group can gain shared understanding while each member's underlying personal space stays separate.
WE ARE RESEARCHING: An open research direction. No availability implied.
PARTNERSHIPS
Six kinds of partner, each with a specific role.
Categories with no publicly approved partner show the invitation instead of a logo. We do not display an organisation unless the collaboration and the wording are approved.
University and research partners
Groups working on distributed systems, personal AI, edge intelligence and governed computing.
Propose a joint research direction or a shared publication. →University of Oulu Future Computing Group(opens in a new tab)
Prifina collaborates with the University of Oulu's Future Computing Group to explore personal and distributed AI, edge intelligence and governed computing across devices, regional infrastructure and cloud environments.
AI model and technology partners
Model providers and capability builders who want to reach people through permissioned access rather than data capture.
Explore how a capability plugs into a personal data space. →Open. We are actively talking to organisations in this category and will name them here once a collaboration and its wording are approved.
Telecommunications and infrastructure partners
Operators and infrastructure providers positioned between devices, regional systems and cloud environments.
Discuss where personal AI workloads should actually run. →Open. We are actively talking to organisations in this category and will name them here once a collaboration and its wording are approved.
Regional and jurisdictional partners
Public sector and regional organisations exploring trusted personal data spaces inside a defined jurisdiction.
Scope a regional personal data space pilot. →Open. We are actively talking to organisations in this category and will name them here once a collaboration and its wording are approved.
Service and shared space builders
Teams building services that need governed access to personal context without owning it.
Build a shared space for a group, network or local area. →Open. We are actively talking to organisations in this category and will name them here once a collaboration and its wording are approved.
Strategic investors
Investors evaluating the personal intelligence layer as a platform category.
Review the thesis, the architecture and the product proof. →Open. We are actively talking to organisations in this category and will name them here once a collaboration and its wording are approved.
NEXT