Services that get you to accurate, production-ready pathology AI—fast.

Pathbox is an AI platform for whole-slide image (WSI) analysis. For labs and biopharma teams, we also offer hands-on services: expert data labeling, model training, and custom model development for detection and analysis.

Start a pilot See services
Full-Service Data & Labeling — We build your ground truth from scratch.
Guided Labeling on Your Data — You define the schema, we execute at scale.
Model Training Accelerator — Hand us your data—get a trained model back.
Custom Model Development — Purpose-built detection and analysis.

Choose your starting point

Train with our pretrained models or use your base models—we support both paths with clear, reproducible pipelines.

Pathbox Pretrained
  • Warm-start with pathology-tuned backbones (segmentation, detection, classification).
  • Fast convergence, strong baselines; great for pilots.
  • Freeze backbone, partial fine-tune, or full fine-tune.
Bring Your Base Model
  • Fine-tune ResNet/ViT/U-Net/Detectron2/SAM variants.
  • Adapters/LoRA options for constrained compute or privacy.
  • Packaged for Pathbox + QuPath or your own inference stack.

Service catalog

Four offerings. Pick one or combine them.

Full-Service Data & Labeling

We provide the data pipeline and the expert labeling team to create a high-quality training set tailored to your goals.

Who it's for: Teams starting from zero or with scattered data.
Inputs we need: Target definitions, inclusion/exclusion criteria.
Timeline: Varies by project scope and complexity.

Deliverables

  • Dataset sourcing/ingest (WSI/tiles), de-identification, and QC.
  • Annotation schema design, pilot labels, consensus guidelines.
  • Expert labeling, adjudication, inter-rater agreement reporting.
  • Exports: WSIs/tiles + JSON/GeoJSON labels + audit logs.
  • Optional: pseudo-labels from pretrained models to boost throughput.

Add-ons

Scanner/LIS integration, S3 storage, viewer access. Pre-label assistance with existing models.

Request a quote

Guided Labeling on Your Data

You bring the WSIs and the requirements; we help you label quickly and consistently at scale.

Who it's for: Labs and biopharma with data in place.
Inputs we need: Schema, examples, acceptance thresholds.
Timeline: Flexible sprint-based approach.

Deliverables

  • Multi-expert labeling with QA and conflict adjudication.
  • Tiled workflows that preserve WSI context.
  • Quality reporting: per-class agreement stats and label consistency checks.
  • Direct handoff into your training pipeline or ours.
Kick off a sprint

Model Training Accelerator

You bring labeled data; we train, tune, and benchmark the model with reproducible experiments.

Who it's for: Teams that want results without devoting internal ML bandwidth.
Inputs we need: Data splits, task definition, KPIs.
Timeline: Depends on data complexity and requirements.

Deliverables

  • Training/validation with experiment tracking.
  • Baseline + ablations; hyperparameter search; early stopping.
  • Starting point: Pathbox pretrained or your base models (ResNet/ViT/U-Net/Detectron2/SAM).
  • Delivery: packaged model (Pathbox/QuPath compatible) + inference API.
Start training

Custom Model Development

Purpose-built detection & analysis for exactly what you need to find—cells, regions, artifacts, or novel patterns.

Who it's for: Labs & biopharma seeking production-grade models.
Inputs we need: Clinical context, annotation samples, target KPIs.
Timeline: Multi-phase approach with pilot and scale-up.

Deliverables

  • Problem scoping and feasibility with success criteria.
  • Model architecture selection (segmentation/classification/detection).
  • Validation plan (holdouts, cross-site, robustness & bias checks).
  • Starting point: Pathbox pretrained or your base models; freeze/partial/full fine-tune; LoRA/adapter options.
  • Packaging: versioned artifact, endpoints, monitoring.

Add-ons

Integration & MLOps (cloud/on-prem), SSO/RBAC, audit logging. Regulatory readiness support (protocols, SOPs, validation docs).

Discuss your use case

How we work

1

Discover

Share your goals, data shape, constraints, and success metrics.

2

Scope

Plan with timeline, deliverables, and milestones.

3

Execute

Label, train, validate—weekly updates and demo checkpoints.

4

Deliver & Iterate

Handover artifacts and integrate feedback.

Ready to get started?

Tell us your use case and timeline—we'll propose the fastest path to a validated result.

Start a pilot