Data handling
On our GPU partner's infrastructure, prompts and outputs are held in memory only while a request is served. Usage records contain the model, token counts, timing and cost.
A secure research platform for universities.
Your research stays protected. Prompts, data and research findings are never shared with model providers for their own use or model training. Your research data and intellectual property remain secure and under your university's control.
Research protection & control
Your university retains ownership of its research data and findings. Sensitive projects run on securely hosted models, without sharing data with external providers.
On our GPU partner's infrastructure, prompts and outputs are held in memory only while a request is served. Usage records contain the model, token counts, timing and cost.
Assign API keys, budgets and access roles by department, lab or research project. Monitor usage and costs through a central dashboard, allocate spending, and configure email alerts for spending thresholds and low credits.
Deployment
Research data, student records and internal documents stay within your perimeter. You decide which workloads, if any, ever leave it.
Start without managing the underlying infrastructure. Divyam.AI handles the cloud deployment.
Access frontier and open-weight AI models through our GPU partner's infrastructure, with a SOC 2 audit currently underway.
Run Divyam.AI in your university's data centre or private cloud. Air-gapped deployment is supported with locally hosted models, keeping research data and findings within your institution's control.
AI spend can follow the budget your institution sets. The knowledge your institution holds can also become the basis for new AI services.
No. Prompts, data and outputs are never shared with model providers for their use, for training or otherwise, and never used to train models for other customers. Our Indian GPU partner never uses them to train or fine-tune any model without your university's prior written consent.
Your university retains ownership of all research data and results.
That depends on the deployment you choose: our managed cloud, partner GPUs in India, or your own campus or cloud. Projects pinned to India-hosted models never escalate overseas.
On our GPU partner's infrastructure, prompts and research findings are held in memory only while requests are processed. Usage records are retained, including the model, token counts, timing and cost. A SOC 2 audit is currently underway with our GPU partner.
Yes. You can run it in your own data centre or cloud account, and air-gapped deployment works with locally hosted models.
Administrators can assign separate API keys, budgets and access roles to departments, labs or research projects. A central dashboard shows model usage, request activity and costs, with spending allocation and email alerts for thresholds or low credits.
Tell us about your research groups, your campus workloads and where your data has to stay.