Choose the AI Mode That Fits Your Research Needs
Three AI modes for different priorities. Use Gemini for speed, combine Gemini and Claude for dual validation, or run everything locally with Ollama for complete data privacy. Your organization, your API keys, your choice.
One-size-fits-all AI does not work for research
Some research teams need the fastest, most affordable AI processing they can get. Others work with sensitive patient data and cannot send a single byte outside their network. And some need the confidence that comes from having two independent AI models validate each other's output.
Most research platforms lock you into a single AI provider. If that provider does not fit your institution's privacy policies, compliance requirements, or budget constraints, you are stuck. You either accept the limitations or look for a different platform entirely.
NousLab solves this with a flexible infrastructure that lets each organization choose its own AI mode. You configure it once at the organization level, and every team member benefits. If your needs change, switch modes without losing a single piece of data.
Different teams, different needs
Three AI modes, one platform
Each mode is optimized for a different priority. Choose the one that fits your organization and switch whenever you need to.
Standard Mode
Gemini for all tasks
Uses Google Gemini for every AI operation: paper analysis, hypothesis generation, protocol writing, and the research assistant. This is the fastest option with the lowest per-token cost, ideal for teams processing large volumes of papers.
- +Fastest processing speed
- +Lowest cost per analysis
- +Single API key to manage
- +Best for high-volume processing
- -Single model validation only
- -Data sent to Google servers
Hybrid Mode
Gemini + Claude for dual validation
Combines two AI providers for different tasks. Gemini handles paper analysis and data extraction where speed matters. Claude handles hypothesis generation and protocol writing where reasoning depth matters. Two independent models reduce the risk of AI blind spots.
- +Dual validation from two models
- +Best model for each task type
- +Reduces AI-specific biases
- +Stronger hypothesis quality
- -Two API keys to manage
- -Higher cost than Standard
Local Mode
Ollama with local models, zero external calls
Runs all AI processing on your own infrastructure using Ollama with open-source models like Mistral and Qwen. No data leaves your network, no external API calls, no third-party processing. Complete data sovereignty for the most sensitive research.
- +Complete data privacy
- +No external API costs
- +Full regulatory compliance
- +Works offline
- -Requires local GPU hardware
- -Slower than cloud models
Your API keys, your costs, full transparency
NousLab never charges you for AI usage. Each organization connects its own API keys and pays providers directly. You always know exactly what you are spending.
Add your API keys
Go to organization settings and enter your API keys for Gemini, Claude, or your Ollama endpoint. Keys are encrypted and stored securely.
Test the connection
Click "Test Connection" to verify your keys work correctly before saving. NousLab validates the endpoint, authentication, and model availability in seconds.
Monitor usage
NousLab tracks every AI call with provider usage statistics. See how many tokens each task consumes, which provider handled each request, and what it cost.
Switch anytime
Change your AI mode or provider at any time from organization settings. All existing data stays intact. Only future AI tasks use the new configuration.
Real scenarios, real flexibility
University hospital using Local mode
A clinical research unit at a university hospital is studying treatment outcomes using patient-derived data. Their institutional review board requires that no patient information leaves the hospital network, not even in anonymized form.
They deploy NousLab with Local mode, running Mistral 7B through Ollama on their own GPU server. All paper analysis, hypothesis generation, and protocol writing happen entirely on-premise. The IT department can verify that zero external API calls are made.
The team gets the same structured workflow as any NousLab user, with full confidence that their sensitive research data never leaves the building.
Pharma company using Hybrid mode
A pharmaceutical R&D team is exploring new therapeutic targets for autoimmune diseases. They need to process hundreds of papers quickly, but they also need high-confidence hypotheses that they can present to their scientific advisory board.
They configure Hybrid mode: Gemini handles the bulk paper analysis at high speed, while Claude generates the hypotheses and research protocols. Having two independent AI models means that each hypothesis has been reasoned through by a different architecture, reducing the risk of systematic AI bias.
The team gets speed where they need it and depth where it matters, all within the same platform and the same project workspace.
Frequently asked questions
What is the difference between Standard, Hybrid, and Local AI modes?
Standard mode uses Gemini for all AI tasks and is the fastest and most cost-effective option. Hybrid mode uses Gemini for paper analysis and Claude for hypothesis and protocol generation, providing dual validation from two different AI providers. Local mode uses Ollama with open-source models like Mistral and Qwen running on your own infrastructure, so no data ever leaves your network.
Do I need to provide my own API keys?
Yes. Each organization manages its own API keys for the AI providers it uses. This means AI costs are billed directly to your account with Google (Gemini), Anthropic (Claude), or run locally at no API cost with Ollama. NousLab provides usage statistics so you can monitor spending and track which tasks consume the most tokens.
Can I switch between AI modes without losing my data?
Yes. You can switch between Standard, Hybrid, and Local modes at any time from your organization settings. All your projects, papers, hypotheses, and analyses remain intact. Only the AI provider used for future tasks changes. Previously generated results are never affected.
Is the Local mode truly private?
Yes. In Local mode, all AI processing runs on your own infrastructure using Ollama with open-source models such as Mistral and Qwen. No data is sent to any external API. This is ideal for research involving sensitive patient data, proprietary compounds, or any scenario where complete data sovereignty is required.
What models are supported in each mode?
Standard mode uses Google Gemini models. Hybrid mode uses Gemini for paper analysis tasks and Anthropic Claude for hypothesis generation and protocol writing. Local mode supports any model available through Ollama, including Mistral, Qwen, Llama, and other open-source models. Custom model endpoints are also supported for organizations with specific requirements.
How do I test my AI configuration before using it?
NousLab includes a test connection feature that verifies your API keys and model endpoints are working correctly before you save the configuration. This ensures your team can start working immediately without troubleshooting connection issues. The test validates authentication, model availability, and response quality in seconds.
Ready to configure your AI infrastructure?
See how NousLab adapts to your organization's privacy, speed, and validation requirements.
Request a Demo