Pathology AI for Everyone
AI Autopilot
Train your own AI pathology model in minutes with a budget of a lunch break. No programming skills required — just prompt your AI agent and go.
Everything You Need to Build Pathology AI
From datasets to inference — every step is automated and managed.
Cohort Creation
Auto-generate training and validation datasets from 66,000+ pre-processed whole slide images.
Model Selection
Pathology-specific algorithms chosen automatically for optimal classification performance.
Training Orchestration
Run training jobs seamlessly on managed cloud GPU compute — no infrastructure setup needed.
Real-Time Monitoring
Track performance metrics, training bounds, and cost live as your model trains.
One-Prompt Deployment
Instantly deploy your trained model securely to the CellDX platform and run inference.
Free Dataset
Access pre-extracted features at no cost. You only pay for the GPU infrastructure.
Agent Compatibility
Use any popular AI agent — Claude Code, OpenAI Codex, Gemini, or open-source alternatives.
No Skills Required
Brief your AI agent with natural language. No coding, no ML expertise necessary.
No Compromise
You choose the AI model architecture you need — full control over your training pipeline.
How It Works
From zero to a deployed custom AI model in a few steps.
Setup
Subscribe to the CellDX platform and create an API key in your account settings.
Install the HistAI SKILLs with your AI agent (e.g., Claude Code).
Train
Prompt your agent with the task — for example, "Train me a Melanoma vs. Nevus classifier."
Confirm the cohort your agent builds, review the cost estimate, and approve the job.
Check performance metrics — if they are not high enough, ask your agent to improve.
Deploy
When the metrics look good, ask your agent to deploy.
Your custom AI model is now trained, deployed, and ready to run.
Run
Upload your WSIs to My Cases and click to run your custom model.
Done — within seconds you receive the model response.
From Months to Minutes
| Traditional Approach | CellDX AI Autopilot |
|---|---|
| Months of development | Minutes |
| $100,000+ cost | $15 – $50 per model |
| Full engineering team | No technical skills needed |
| Complex infrastructure | Fully managed |
| Manual iteration | Automated optimization |
The Platform in Action
Pay Only for Compute
Access to our 66,000+ slide dataset is free of charge. You only pay for the GPU time you actually use.
No Infrastructure Setup
Fully managed platform — no GPUs to provision, no pipelines to configure.
No Engineering Required
Prompt-based interface powered by AI agents you already use.
The Engine Under the Hood
Features are pre-extracted from every slide using state-of-the-art proprietary foundation models. Training operates on compact representations — that is why it takes 30 minutes, not 30 hours.
Data Layer
Whole slide images with pre-extracted features stored in optimized sharded storage.
Intelligence Layer
Automated model selection, parameter optimization, and cross-validation.
Application Layer
Self-service UI, real-time dashboard, and one-prompt deployment.
The Future of Pathology AI
Is Self-Service
Every pathologist has questions AI can answer at scale. AI Autopilot gives them the power to build the answers themselves.
slides ready today
average model creation
fully-managed end-to-end pathology AI platform