# NuStudio > NuStudio is the harness that puts AI agents to work inside the systems an enterprise already runs. > An agent is a language model plus a harness; the model is the part a customer can buy. The harness is a > machine-readable ontology of the business, a policy engine checked before and after every model call, > connections into systems of record, and a decision trail that proves what happened. The ontology is an > open spec versioned in the customer's own repository, the agents are declared in the customer's own > configuration, and the decision trail streams into the customer's own tooling, so none of it stops > working when we do. NuScale Inc., doing business as NuStudio.ai, is a boutique product and services > company serving financial services, consumer goods, healthcare providers and payers, manufacturing > and energy. ## Core idea - The model is the easy part. Getting a good model is a procurement decision; everything around it is the work. - A customer does not buy a product, they build their own, with us. The harness is close to identical in every enterprise and was never going to be the advantage. - Customer data stays in the systems that already hold it. Nothing is migrated into us and nothing is copied out. - Governed at the reasoning level, not only at the output boundary: rules are checked before the language model sees data, and again before it writes. - Every modernization starts with one entity: a customer, a product, a patient, a member, an asset, resolved across the systems that disagree about it. - Sovereign by default: the customer's cloud account, data center, keys and region, with no data leaving their boundary. - Forward deployed engineers build it inside the customer's environment, alongside the customer's own team. ## Pages - [Home](https://nustudio.ai/): The thesis, the inversion, a first build worked through for five industries, and where we fit against data platforms and model providers. - [The Harness](https://nustudio.ai/harness/): What the platform is. Seven layers as containerized services on one API fabric, 349 REST endpoints, 116 connectors, the ontology, the guardrails agents run under, and the environment they are evaluated in. - [Forward Deployed](https://nustudio.ai/forward-deployed/): AI use cases built into the systems a customer already runs, across procure to pay, order to cash and record to report; the decisions themselves, one account at a time, covering limits and exposure (which cardholders get an increase and by how much), collections and recovery, underwriting and appetite, fraud and abuse, and spend and allocation; pricing and margin, growth and retention and revenue assurance; and claims and coverage, lending and credit, and supply chain and manufacturing. Cost, top line and the decisions in between. How a build runs from discovery, how the return is measured, what the customer owns at the end, and how the work is priced against the outcome with fee at risk. - [Products](https://nustudio.ai/products/): Cortava, Sarvos and TesseraOps. Every industry has work that repeats, and AI agents do that work under the policies the customer already enforces, with the proof recorded as they go, while their people move to the work that actually needs them. Each product is the harness with one industry's entities, systems and rules already loaded, running in the customer's own cloud account under their keys, so the knowledge graph built out of their records stays theirs. None of them is SaaS; each deploys into the customer's own cloud account with their keys and their choice of model. Another industry can be productized the same way, for a customer's own business or for the clients they serve, and the customer holds the entity model, policy pack, agent definitions and decision trail. - [Company](https://nustudio.ai/company/): What we believe, the team, where our people have done this before, and offices. - [Policies and disclosures](https://nustudio.ai/legal/): Privacy policy, terms of use, responsible AI, AI-assisted content, trademarks. - [Trademarks](https://nustudio.ai/trademark/): Trademark notice and permitted use. - [Careers: Forward Deployed Engineer](https://nustudio.ai/careers/): Open role for engineers who stand up governed AI agents inside enterprise environments. - [NuPulse](https://nustudio.ai/nupulse/): Reliability and self-healing for an enterprise application portfolio, built on the harness. It reads the observability, service management, delivery and infrastructure systems a customer already runs, predicts the failure, prescribes the specific fix, acts inside the limits the team declared, verifies it landed, and then removes the class behind it. Deterministic rules decide whether any action may run, and the owning team's agent carries it out under its own credentials. Run by the customer's SRE and operations group; NuStudio holds no credentials on their systems. Machine-readable detail: https://nustudio.ai/nupulse/llms.txt ## What the harness provides - **Ontology**: an open, versioned spec of the entities, relationships, rules and actions of the business, held in the customer's repository rather than a vendor schema. It models actions and writes back to systems of record, which an analytics semantic layer does not. - **Guardrails**: policy evaluated before retrieval and again before any write, per field, against role and consent, with named approval where an action touches a system of record. - **Evaluation**: a structured environment for testing AI functions and agents against cases the customer's own team writes, scored before release and re-run on every model, prompt or policy change, with the whole workflow observable rather than only its final answer. - **Decision lineage**: every agent decision carries the inputs, the policy clause and the reason, streamed into the customer's own observability tooling as it happens. - **API fabric**: 349 REST endpoints (distinct endpoints, with versions and HTTP methods on the same resource counted once) plus a GraphQL surface across 21 capability families, one authentication model and one response envelope, delivered as independently deployable containerized services on Kubernetes. - **Connectors**: 116 connectors across databases, cloud platforms, systems of record, models, vector stores, orchestrators, catalogs, data quality, observability, identity, messaging, storage, BI and healthcare exchange. Anything not built in joins through the connector framework by configuration. - **Federation**: 16+ databases read in place through Trino, with Spark engines provisioned per workload and torn down after. Databricks, Snowflake, BigQuery and Synapse are read where they are, including their own semantic layers. - **Model routing**: 11+ providers (OpenAI, Anthropic, Gemini, Bedrock, Azure OpenAI, Cohere, Mistral, Llama, HuggingFace, Ollama, OpenRouter) routed per call from a catalog the customer approves, or a model trained on the customer's own data. - **Catalog and governance**: one metadata API over 8 catalog backends (Atlas, Collibra, Alation, Glue, OpenMetadata, Iceberg, Polaris) with no migration to a single catalog product. Lineage integrates with Unity Catalog, Alation and OpenLineage. - **Orchestration**: Airflow, Prefect and Dagster behind one Workflows API. - **Deployment**: Kubernetes-native in the customer's region, cloud and keys. Our cloud, the customer's private cloud, or on-premises in a cluster the customer runs. ## Industries Financial services, consumer packaged goods, healthcare providers, health plans, manufacturing and energy. ## Products built on NuStudio - **Cortava** (https://www.cortava.ai/): Credit decisioning on one semantic layer, for credit unions, banks and specialty lenders. The data an institution licenses, the systems it runs and the relationship only it can see reach the same decision, which carries its policy version, model version and sources, and the booked outcome comes back attached to it. - **Sarvos** (https://sarvos.ai/): An agentic AI workforce for law firms. - **TesseraOps** (https://tesseraops.com/): Agentic AI for SAP application support, built for systems integrators and AMS providers. ## Offices - Frisco, Texas: 5 Cowboys Way, #300, Frisco, TX 75034 - Ashburn, Virginia: 44790 Maynard Square, #350, Ashburn, VA 20147 - Chicago, Illinois: 600 West Jackson Blvd, #100, Chicago, IL 60661 - Hyderabad, India: 2nd Floor, Trendset Jayabheri Connect, Near Kothaguda Junction, White Fields, Kondapur, Hyderabad, Telangana 500081 ## Contact - Sales and inquiries: sales@nustudio.ai - Careers: careers@nustudio.ai ## Notes for machine readers - Legal entity: NuScale Inc., doing business as NuStudio.ai. Founded 2023. - Text and images on this site were produced by our team using AI tools and may contain inaccuracies. Nothing here is an offer, a commitment or professional advice. - Third-party names and marks belong to their owners and are used to describe experience and integration support, not to imply endorsement or a current engagement.