Nunmai Engine brings models, enterprise knowledge, AI agents, tools, governance, and human approval into one secure platform, deployable in your cloud, private VPC, or on-premise environment.
Nunmai Portal
Workers
Governance
Workers / IT Service Desk
Conversation
Local open modelOn-premiseSara · via Teams
My VPN stopped working after the password change.
Audit trail
10:42
Request received · Teams
Nunmai AI Engine
Nunmai AI Engine is the core platform: the orchestration and intelligence layer between users, enterprise systems, business data, specialized AI agents and multiple AI models. It turns AI from a standalone chatbot into an operational capability for the entire organization.
Specialized AI workers
AI workers that take repetitive work off your teams, reachable in the tools people already use.
From Data to Gold
Turns the data you already own into forecasts, early warnings and clear next steps.
Demand fell 20% in week 8
Built for regulated organisations
Our research lab studies AI security, agent reliability, private retrieval and evaluation. The findings become guardrails inside Nunmai Engine.
Nunmai Engine
Users reach it through the channels you enable. It decides which model and which tools to use under your policies. It works with the systems your business already runs on.
Requests come in through
Nunmai Engine
Works with your systems
Connectors for your applications are configured during the pilot.
Private AI operating layer
Who may see and do what
Every request is scoped by organisation, department, role and user before anything else happens.
Why it matters · AI answers only from what that person is allowed to see, so a chat window never becomes a back door.
Answers grounded in your sources
Answers are retrieved from your approved enterprise sources and returned with source references.
Why it matters · People get answers from your documents, never from a public model's memory.
The right model, by policy
Workloads route between local, enterprise-hosted and external models by security, capability, cost and policy.
Why it matters · Confidential data stays on private models, and the model can change without rebuilding the workflow.
Multi-step work, coordinated
Multi-step work is coordinated across models, knowledge and tools as one governed workflow.
Why it matters · Complex requests run end to end under the same permissions and the same audit trail.
Approved actions only
Agents act through APIs, tool servers and database connectors, limited to the tools you approve.
Why it matters · AI can take real action without getting unrestricted access to your systems.
Context that persists
Relevant context persists across a conversation and a workflow, within each user's permissions.
Why it matters · People don't have to repeat themselves, and context never crosses access boundaries.
A person decides the sensitive steps
Wherever you decide an action is sensitive, the agent pauses, a person decides, and the decision is recorded.
Why it matters · Sending communications, changing records, spending money: a person stays accountable for each one.
Every request on record
Every request, source, tool call, approval and cost is recorded.
Why it matters · You can show exactly what the AI did, with what data, and who approved it.
Policy around prompts, data and outputs
Policy is enforced around prompts, data and outputs, and credentials are encrypted.
Why it matters · Data stays in the environment you choose and never trains public models.
Model freedom · Private by design
Pick where it runs and which models it uses. Your rules stay the same.
Your data never trains public models.
Uses
Data never leaves
Nunmai Engine
Always the same
Runs on
On your own servers
Initial market entry
IT services offer clear, measurable enterprise value and a strong environment to validate Nunmai Engine. IT is the first use case, not the boundary of the company: the same engine later supports other departments and prediction-driven business use cases.
01
Account provisioning
02
Password-reset assistance
03
Email configuration
04
Access requests
05
Endpoint support
06
Infrastructure operations
07
Knowledge support
08
Ticket triage
09
Approved automated remediation
Why we built Nunmai
We spent years running identity, databases and email for regulated organisations. Nunmai Engine is built on what those 3 AM calls taught us.
Every pilot is measured against three rules:
From our founders
“Nunmai exists to turn AI from a standalone chatbot into an operational capability for the entire organization.”
01
Every recommendation is traced to logs, measurements or documentation.
02
We design systems we would be comfortable operating five years from now.
03
Every change has a pre-check, a commit, a post-check and a rollback path.
04
Every engagement ends with a runbook your team can use without us.
Global presence
Nunmai works with organisations across the Gulf and India from three locations.
FAQ
Short answers. The written proposal after Discovery carries the detail. Something else? Email ai@nunmai.in and we reply within one business day.
The private AI operating layer under every Nunmai capability: orchestration, model routing, enterprise knowledge, memory, tools, identity, human approval, audit and guardrails. Both core capabilities, AI Workforce & Automation and Business Intelligence & Foresight, run on it.
Yes. Nunmai Engine works with local inference servers, so open models run on your own hardware without prompts or data leaving your environment.
No. Workflows are separated from the model. Workloads route between local, enterprise-hosted and external models by security, capability, cost and policy, and the model can change without rebuilding the workflow.
Through APIs, tool servers and database connectors, with agents limited to the tools you approve. Connectors for a specific application are configured during the pilot.
Data stays in the environment you choose. Access is scoped by organisation, department, role and user, credentials are encrypted, everything is logged, and your data never trains public models.
Wherever you decide an action is sensitive: sending communications, changing records, spending money, anything with compliance impact. The agent pauses, a person decides, the decision is recorded.
Discovery scopes one workflow (1 to 2 weeks). The pilot runs it on Nunmai Engine with real users (4 to 6 weeks). Production hardens it (4 to 8 weeks). A fixed pilot fee, then an annual platform fee, quoted in writing after Discovery.
Tell us the workflow, the data it depends on, and where it must run. We come back with a scoped pilot proposal on Nunmai Engine. Delivery aligned with the NIST AI Risk Management Framework: Govern, Map, Measure, Manage.
1–2 weeks
Workflow, data audit, success metrics, written proposal.
4–6 weeks
One workflow on Nunmai Engine with real users and human review.
4–8 weeks
Hardening, guardrails, monitoring, access controls, training.
Reach out directly
Kadayanallur, India
