Demo listing. This page uses illustrative seed metadata to demonstrate the Atlas. Claims shown here are not verified product facts.
Chainspan
Mixture-distance service that computes Chain Rule Optimal Transport (CROT) metrics between statistical mixtures — giving ML and ranking…
A mixture-distance service that computes Chain Rule Optimal Transport (CROT) metrics between statistical mixtures — giving ML and ranking teams tractable, metric distances and tight upper bounds where KL and f-divergences have no closed form.
Last verified 15 Sept 2026 · Updated 23 Sept 2026 · Metadata v4
Problem
Platform ML teams that compare, cluster, or monitor Gaussian/other mixtures (user embedding mixtures, topic mixtures, KDE audience models) and today rely on expensive Monte Carlo f-divergence estimates.
Outcome
A design-complete foundation in the active pipeline — engineering scaffold planned to complete this month.
Who uses it
- Operations Analyst
- Solution Architect
Key capabilities
Capabilities are defined during customer configuration.
Typical workflow
Primary operating flow
Trigger: An operator starts the primary use case
Task
Stage 1: Capture inputs
Operator
AI-assisted
Stage 2: Assist with draft
System
Human checkpoint
Stage 3: Human approval
Approver
Task Automated AI-assisted Human checkpoint Decision
About this foundation
Statistical geometry API for mixtures. Information-fusion and ranking stacks constantly need dissimilarities between mixtures; KL/f-divergences lack closed forms and need costly MC. Chainspan commercialises CROT: optimal transport on marginals with ground distance on conditionals (chain-rule factorisation), proven metric when the ground distance is metric, generalising Wasserstein and mixture-component OT (MCOT), with Sinkhorn relaxation and documented upper bounds on TV, √JS-style, Wp, and Rényi distances under joint convexity.
Related applications
- Fieldproof
Deployment assurance system for funders and implementing organisations putting AI into high-stakes humanitarian and development work:…
- • Addresses the same problem: Manual Workflow, Fragmented Systems
- • Shared capabilities: Workflow Automation, Document Intelligence
- Revaloop
Disposition and residual-value system for durable-goods producers that decides, unit by unit, which next cycle a returned product should…
- • Addresses the same problem: Manual Workflow, Fragmented Systems
- • Shared capabilities: Workflow Automation, Document Intelligence
- Benchline
Capability-runway planner for large employers that forecasts skill demand only as far out as the slowest supply lever can respond, then…
- • Addresses the same problem: Manual Workflow, Fragmented Systems
- • Shared capabilities: Workflow Automation, Document Intelligence
- Centaura
Decision-rights and override ledger for ai-assisted work that grades each use case's earned autonomy by what it can explain and prove,…
- • Addresses the same problem: Manual Workflow, Fragmented Systems
- • Shared capabilities: Workflow Automation, Document Intelligence
- Curvebank
Reference-class forecasting desk for new-technology ramps that forecasts a launch by naming the historical ramp it will resemble,…
- • Addresses the same problem: Manual Workflow, Fragmented Systems
- • Shared capabilities: Workflow Automation, Document Intelligence
- Embedra
Representation registry for systematic trading research that treats the signal-compression layer — not the classifier — as the governed,…
- • Addresses the same problem: Manual Workflow, Fragmented Systems
- • Shared capabilities: Workflow Automation, Document Intelligence
Primary operating flow
Trigger: An operator starts the primary use case
Task
Stage 1: Capture inputs
Operator
AI-assisted
Stage 2: Assist with draft
System
Human checkpoint
Stage 3: Human approval
Approver
Actors
Operator, Approver
Outputs
- Completed work item
Evidence generated
- Audit trail
Exceptions
- Incomplete inputs returned to operator
Task Automated AI-assisted Human checkpoint Decision
Included ships in the foundation · Configurable is switched or tuned per customer · Extension is customer-specific build scope · Planned is on the roadmap and not available today.
Core workflow
AI-assistedPlannedPrimary operating workflow for this foundation.
Document intelligence
AI-assistedPlannedExtract and ground decisions from operating documents.
- Architecture class
- Domain-driven design, OpenAPI-first
- Bounded contexts
- 3
- API-first
- Yes — OpenAPI contracts are authoritative
- Identity
- Enterprise OIDC / SAML via the fazeZERO identity blueprint; tenant-aware role-based access.
- Multi-tenancy
- Tenant-aware
- Service boundaries
- Bounded contexts behind one API server.
- Integration approach
- Ports-and-adapters: every external system sits behind an adapter; OpenAPI contracts for inbound APIs.
- Eventing
- Integration events via transactional outbox.
- Storage abstraction
- Repository interfaces; DynamoDB or relational adapters.
- Deployment pattern
- Containerized API + web application; infrastructure as code per cloud profile.
- Architecture version
- 1.0
Approved domain names
- identity
- catalog
- workflows
- Web Application
- API Server
- Application Services
- Domain / Generated Core
- Adapters
- Customer Systems
Generator source, templates and factory orchestration are proprietary and are not part of this listing.
Reference Adapter: shipped and tested · Previously Integrated: delivered before · Standard API Pattern: integrates via a documented pattern · Customer-Specific: built in your implementation.
APIs
| Integration | Status | Direction | Method |
|---|---|---|---|
| Customer systems of record | Standard API Pattern | bidirectional | REST API |
Deployment
- Deployment models
- Customer Cloud, Private Cloud
- Cloud profiles
- Azure Profile, Cloud-Neutral
- Containerized
- Yes
- Regions
- GLOBAL
Cloud profiles describe approved deployment patterns. They are not formal marketplace certifications.
Data handling
- Stores customer data
- Configurable
- Data leaves customer environment
- Configurable
- Uses external AI provider
- Configurable
- Sends logs externally
- No
- PII expected
- Determined during customer configuration
- Data-residency support
- Yes
- Subprocessors required
- Determined during customer configuration
Identity & access
- Authentication
- OIDC, Entra ID, SAML
- Authorization
- Role-Based, Tenant-Aware
- Source availability
- Proprietary
Support
- Implementation
- Delivered by fazeZERO or a certified partner during the Solution Definition and AI Production Sprints.
- Production support
- Production support available under a separate support agreement.
- Contact
- support@fazezero.com
Badges
fazeZERO Verified
Listing metadata and published technical claims have been reviewed by fazeZERO.
Issued 23 Sept 2026
Evidence shows that this listing represents real engineering. Internal metrics are only published when fazeZERO has approved them for publication.
Build
Automated tests
Deployment validation
- Last verification
- 15 Sept 2026
- Release
- 1.0.0
- Architecture version
- 1.0
- Listing metadata version
- 4
- Last published
- 23 Sept 2026