Foundation Models
Frontier and open-weight models behind one API contract.
Platform
The Enterprise Intelligence Platform
One control plane across models, agents, knowledge, data and compute — so the sixth use case lands where the first one did, with the same audit trail behind it.
One environment
Not a toolkit. An operating environment you keep.
Models, data connections, deployments, guardrails, compute and the evidence behind every decision — in one console, inside your own tenancy.
The sixth use case lands where the first one did
One control plane, one audit trail, one access model. The alternative is six vendors, six consoles and six governance stories to defend separately.
Everything the console does, the API does
Projects, registry entries, deployments, guardrails and evaluators are all addressable, so the estate is manageable as code rather than by browser.
BasinWrightOverview
Modules
Model 360
drift · evidence
Deployments
endpoints · rollback
GPU 360
fleet · utilisation
Data Hub
stores · queries
Guardrails
screening · evaluators
Automations
workflows · triggers
What is in it
Everything an AI estate needs, in one operating environment
Instead of assembling model vendors, GPU providers, a vector store, an evaluation harness and a governance story, you operate one environment — and keep it.
GPU Cloud
H100, H200 and B200 capacity with multi-node interconnect.
Agent Platform
Autonomous agents with memory, planning and tool calling.
Enterprise Search
Semantic retrieval across every system your business runs on.
Knowledge Hub
Governed, versioned corpora with lineage back to the source.
Security
Private networking, encryption in transit and at rest, BYOK.
Governance
Policy, audit trails and evaluation gates before promotion.
Fine-Tuning
LoRA and full-parameter tuning on your own data, in your tenancy.
Model Marketplace
Models, agents and extensions with commercial terms attached.
Deployment
Serverless, dedicated, on-premises or fully air-gapped.
Models
Deterministic and generative, deliberately both
A lot of vendors have quietly reduced every problem to a prompt. That is the wrong tool for most enterprise decisions, and it is why so many pilots die in risk review. Route by cost, latency, residency or capability without rewriting application code.
model
Deterministic ML
Scorecards, anomaly detectors, survival and time-series models — reproducible and defensible.
model
Large Language Models
Reasoning, drafting and synthesis, self-hosted and fine-tuned on your corpus.
model
Vision Models
Detection, inspection and document understanding.
model
Speech Models
Transcription, diarisation and real-time voice.
model
Embedding Models
Dense and sparse retrieval tuned for your corpus.
model
Multimodal Models
Text, image, audio and video in a single context.
model
Code Models
Generation, migration and review across your stack.
model
Domain Specific Models
Clinical, legal, financial and industrial specialists.
Supported providers
OpenAIAnthropicMetaMistralGoogleDeepSeekQwenMicrosoftCohereOpen Source
Catalogue
A catalogue to start from, not a blank page
A shipping use case arrives as a working data pipeline, a trained artifact, an evaluation report and a model card. It is a starting point: what runs in your production is calibrated on your population, your policy and your thresholds.
Tuned on your data, always
The reference model is where the work starts. Deterministic models are recalibrated and re-validated; LLM components are fine-tuned, grounded and guardrailed on your corpus.
Build to order on the same pipeline
What is not in the catalogue is specified and built the same way, with the same governance and the same monitoring — a productised service rather than a research project.
BasinWrightModels & Registry
Catalogue
tune to your dataPromptShield Injection Detector
AI security
LeakSentry Data Leakage Detector
AI security
GovernanceLens Compliance Scorer
EU AI Act · NIST AI RMF · ISO 42001
RAGArmor Vulnerability Scanner
Knowledge & GenAI
PoisonGuard Training Data Screener
AI security
Not in the catalogue
specified, then built on the same pipeline
Agents
A digital workforce that knows how your business runs
Agents ship with memory, planning, reasoning, tool calling, workflow automation, human approval and knowledge retrieval — and escalate when policy requires it rather than improvising.
Procurement Agent
Screens suppliers, drafts RFPs and flags contract drift.
Finance Agent
Reconciles ledgers, explains variance and drafts board packs.
HR Agent
Answers policy questions and shepherds onboarding end to end.
Customer Service Agent
Resolves tier-one volume with full case history in context.
Legal Agent
Reviews clauses against playbooks and surfaces obligations.
Supply Chain Agent
Watches signals upstream and re-plans around disruption.
Sales Agent
Researches accounts, drafts outreach and keeps CRM honest.
Engineering Agent
Triages incidents, proposes fixes and writes the postmortem.
Model 360
An outcome you cannot measure is a claim
Every model carries the same evidence, on by default: why it decided what it did, a signed trail back to the inputs and sources behind it, and continuous monitoring of whether it still holds.
Never a bare number
Contributing factors and reason codes on tabular models, cited source passages on retrieval and LLM models. Adverse-action reason codes are native, which matters wherever a declined applicant has a legal right to an explanation.
Signed reasoning receipts
A cryptographically signed evidence trail linking an output back to its inputs and its sources — built for audit and for dispute resolution rather than for a dashboard.
Watched, not assumed
Drift and accuracy decay monitored statistically, output quality scored by evaluators continuously rather than once at UAT, and every inference logged and metered.
BasinWrightModel 360
Evidence on every decision
Reason codes & contributing factors
tabular models
Cited source passages
retrieval & LLM models
Signed reasoning receipt
Ed25519 · output linked to its inputs
Data-to-decision lineage
versioned config · change history
Drift & accuracy decay
PSI / CSI · statistical monitoring
Evaluators
rubric scoring, continuous
Serving
Runs where your policy allows it to run
Residency, sovereignty and blast radius are usually decided before the model is. The same platform, the same governance and the same evidence trail follow the model to whichever of the three it lands on.
Not everything needs a GPU
CPU serving covers a large share of deterministic use cases at a fraction of the cost, and we will say so when a use case does not need the hardware.
Your data does not move to reach it
The platform database holds application state and metadata. Your business data stays in your systems and is read at query time — an architectural invariant enforced in the codebase, not a policy written for a website.
BasinWrightDeployments
Serving targets
CPU serving
inside your environment
scoring · forecasting · classification
Cloud GPU
provisioned on demand
embeddings · document intelligence · LLM inference
Your own GPU
on-premises, inside your perimeter
sovereignty & residency-constrained work
versioned · logged · rollback-ready
For engineers
Driven by API, not just by console
A public REST API across the full platform surface, SDKs for Python and TypeScript, an OpenAI-compatible inference gateway that authenticates and meters every call, and MCP support so your models and data are reachable as tools by agentic clients.
Anything you can click, you can automate. Projects, registry entries, deployments, guardrails and evaluators are all addressable, which is what makes the platform survivable in an organisation that manages infrastructure as code rather than through a browser.
Every call routes through the gateway, so usage, cost and behaviour are visible per project, per model and per team — the same telemetry that makes drift monitoring possible in the first place.
Modules
What you operate day to day
Every one of these is a surface in the console and an endpoint on the API. Nothing here is on a roadmap.
Projects
Isolated workspaces with their own data, credentials, compute and members — the unit of separation between teams, environments and business lines.
Data Hub
Register, browse, preview and query the data stores your models read from.
Integrations
Connect databases, warehouses, object stores, streams and AI providers. Credentials sealed and masked at rest.
Models & Registry
Catalogue, version, configure and promote models through their lifecycle, with evaluation gates before promotion.
Deployments
Live endpoints with health, logs, events and rollback per deployment.
Services & Templates
Provision the supporting infrastructure — search, vector and query engines — from templates onto registered servers.
GPU 360
Fleet inventory, live utilisation and cost per GPU server, across cloud capacity and your own hardware.
Chat & Knowledge
Grounded conversational access to your own documents and data, with the passage an answer came from attached.
Guardrails & Evaluators
Safety screening on inputs and outputs, plus rubric-based quality scoring run continuously rather than once at UAT.
Automations
Multi-step workflows chaining models, data and actions on a schedule or a trigger.
Access Control
Role-based and attribute-based permissions with deny-overrides, down to the individual resource.
Billing & Usage
Per-project consumption, rate cards, credits and spend visibility — because every inference routes through a gateway that meters it.
Talk to us
See it against your own estate
The useful version of a platform demo is one pointed at your systems and your constraints. Tell us what those are and we will run it that way.