Data sovereignty
Contracts, HR files, designs and customer records stay on servers you own.
We build the systems other companies run on.
Enterprise AI, spatial and immersive platforms, mobile products, and custom solutioning — architected, built, and handed over by one team. Now with private AI that runs on your own hardware.
The idea
Grid is the infrastructure — ordered, engineered, dependable. Nova is what happens when that structure comes alive.
New Gridnova On-prem AI
Your AI. Your hardware. Your data stays put.
We size, install, and run a private AI stack inside your building — so no prompt, file, or record ever leaves your network. Try it.
1 · Choose how isolated it is
Most chosen
2 · Size it for your company
Indicative setup
Department
Send this setup to our teamIndicative only. Final sizing follows a short assessment.
Drag to look around · tap a number
0questions answered on-site
0 bytessent outside your network
Lockedinternet link
3 · Ask your private assistant simulation
What we set up
GPU server specification matched to your users and workloads — procured, racked and configured.
Open-weight LLMs selected, benchmarked on your tasks and served locally.
Your files indexed into a private vector store, with answers that cite their sources.
MCP servers and integrations into ERP, CRM, file shares and databases — read-only by default.
A branded assistant for staff, with SSO, roles, per-department knowledge and full audit logs.
Monitoring, model updates, re-indexing and an annual support plan, run by the team that built it.
Why on-prem
Predictable cost
Hardware and support you budget for once, not per-seat or per-token bills that grow with adoption.
Grounded in your knowledge
Answers come from your documents and databases, with sources cited.
No vendor lock-in
Open-weight models you can swap as better ones ship. Your AI keeps running whatever a provider decides.
Rollout
- AssessUse cases, data sources, users and security rules.
- SizeHardware and model shortlist, benchmarked on your tasks.
- DeployInstall, harden, connect SSO and audit logging.
- GroundIndex documents, wire connectors, tune answers.
- OperateTrain your teams, monitor, update and support.
Questions buyers ask
Is a private model as good as the big cloud assistants?
For work grounded in your own documents — policies, contracts, manuals, records — a well-chosen open-weight model with good retrieval is very capable. We benchmark candidates on your real tasks before you commit.
Do we need our own data centre?
No. Smaller setups run on a single server in a comms room. Larger ones go into your existing rack or private data centre.
Who maintains it after go-live?
We do, under an annual support plan — or we train your IT team to run it, with runbooks handed over either way.
Can it connect to our existing systems?
Yes. We build connectors to ERP, CRM, file shares and databases using Model Context Protocol servers and standard APIs, read-only unless you decide otherwise.
Can departments see different information?
Yes. Access follows your directory groups, so HR documents answer only for HR, finance for finance, and every query is logged.
Who we are
A full-stack partner for systems that have to hold up
Gridnova is a full-stack software agency built around B2B delivery. We design and ship scalable web, mobile, desktop, and backend applications — from complex enterprise integrations and production AI systems to design-intensive interactive applications.
One team owns the architecture, the build, and the handover. Nothing gets thrown over a wall.
What that buys you
One architecture, one accountable team, and a system your engineers can still run after we leave.
New
On-prem AI
Private AI stacks sized, installed and supported on your own hardware.
Enterprise applications
Web, mobile, and desktop products built for scale and long service life.
AI & RAG platforms
Retrieval systems and MCP servers grounded in your proprietary data.
3D, spatial & AR
Browser-native visualisation, AR overlays, and indoor wayfinding.
Architecture & backend
Microservices, integrations, and the scalability work underneath.
Where the work has gone
Six domains, one engineering standard. We take work where the requirements are hard and the tolerance for failure is low — factory floors, construction models, delivery networks, and enterprise data.
0production systems delivered
0industry domains served
0median time to first production release
0uptime across deployments we operate
Domain expertise
Six domains, six different hard problems
01 · Advanced AI platforms
Intelligent systems grounded in your own data
We build AI that answers from your documents and databases — not from the open internet.
- Designed and deployed scalable AI platforms tailored to B2B workflows, with retrieval, evaluation, and access control built in from the start.
- Built Retrieval-Augmented Generation systems connecting large language models to internal databases and document stores.
- Developed Model Context Protocol servers to standardise secure AI-to-datasource communication across client systems.
RAGMCP serversPythonPostgreSQLVector search
02 · Construction tech & BIM
Full building models, running in a browser tab
BIM software that removes the desktop licence from the review loop.
- Built applications that handle industry-standard BIM formats — IFC for open exchange, plus Revit (.RVT), OBJ, FBX, and STL.
- Developed web-based 3D visualisation in Three.js, enabling interactive model inspection, navigation, and data overlays directly in the browser.
- Integrated BIM workflows with backend services for secure model storage, versioning, and metadata management.
IFCRevit / RVTThree.jsOBJ · FBX · STL
03 · Manufacturing automation
Shop-floor documentation that keeps itself current
Real-time digital oversight replacing manual, error-prone tracking.
- Engineered a system to track manufacturing processes, manage workflows, and automate documentation end to end.
- Replaced manual paper tracking with real-time oversight, giving management live visibility into every line.
- Designed interfaces for factory-floor conditions — gloved hands, poor light, short attention windows — alongside analytics dashboards for management.
ReactNode.jsPostgreSQLReal-time sync
04 · Pharmaceutical manufacturing
Every batch, visible from the first step to the last
PDS — a production management application for a pharmaceutical company's manufacturing unit.
- Built PDS, a production management application for the manufacturing unit of a pharmaceutical company.
- Follows each medicine from the moment its batch is created, through every stage of production.
- Gives the plant one clear view of the whole process — which batch is at which stage, and what comes next.
Batch trackingStage-by-stage workflowProduction visibility
05 · E-commerce logistics
Scheduling and routing at delivery-network scale
A platform that decides what goes where, and in what order, before the vans move.
- Built order allocation, delivery planning, and route optimisation for a high-volume e-commerce operation.
- Owned the real-time data models and backend services underpinning dynamic scheduling workflows.
- Improved operational efficiency through automated resource allocation, holding delivery windows under peak load.
In production
15korders scheduled per day
90 min → 4 minroute planning cycle
PythonNode.jsRoute optimisationMicroservices
06 · Interactive, AR & wayfinding
AR product showcases and indoor wayfinding
Spatial work for brands that want their product in the customer's room, and for buildings people struggle to navigate.
- Built AR lenses for consumer brands — product showcases that let a customer place, scale, and inspect an item in their own space before they buy.
- Developed interactive wayfinding for complex indoor environments such as hospitals, campuses, and retail, optimising routing and the first-time-visitor experience.
- Optimised the rendering pipeline around the device, not the demo, so spatial interactions hold frame rate on the phones customers actually carry.
ARKit / ARCoreAR lens platformsThree.jsSwiftKotlinReact Native
01 / 06
A delivery model you can plan a quarter around
Every engagement runs the same five phases. You know what happens next, and what you get at the end of each one.
- 01
Discovery
Requirements, constraints, and success criteria, written down and agreed.
1–2 weeks - 02
Architecture
System design, data model, integration map, and a delivery plan with dates.
1–2 weeks - 03
Build
Two-week sprints with a working demo at the end of each one.
Sprint cadence - 04
Hardening
Load testing, security review, and performance work against agreed targets.
2–3 weeks - 05
Handover
Documentation, runbooks, and your team walked through the system.
Included
What high standard actually means here
These are commitments, not aspirations. They apply to every engagement.
Review on every merge
No code reaches main without a second engineer reading it.
Automated pipelines
Build, test, and deploy run on every change, not on demand.
Performance budgets
Frame rate and latency targets agreed up front and tested against.
Security by default
Access control, secrets handling, and data boundaries designed in, not retrofitted.
Documentation as deliverable
Architecture notes and runbooks ship with the code.
Clean handover
Your team can run, extend, and debug the system without us.
The stack behind our solutions
Chosen for performance, security, and cross-platform reach — not novelty. We pick the boring, well-supported option unless the problem genuinely requires otherwise — and we tell you which one we picked and why.
Product surface
- React, TypeScript, JavaScript
- Native iOS (Swift), Android (Kotlin)
- React Native for cross-platform
- Electron for desktop
Systems & data
- Python, Node.js
- REST APIs, microservices
- PostgreSQL, MySQL, MongoDB
- Enterprise system integrations
AI & spatial
- RAG pipelines and evaluation
- Model Context Protocol servers
- LLM integration and grounding
- Three.js, modern 3D processing
On-prem AI
- Open-weight LLMs, served locally
- GPU inference servers
- Private vector databases
- SSO, RBAC and audit logging
Why teams bring us the difficult half
True full-stack capability
Frontend, mobile, desktop, and backend under one roof — so architectural decisions get made once, by people who will live with them.
Cross-domain intelligence
AI, BIM, manufacturing, pharma, logistics, and spatial computing. Patterns from one domain routinely solve problems in another.
Future-ready architecture
Systems designed to be extended by whoever comes next, not just to pass acceptance.
Business-centric approach
We start from the operational problem and the number that has to move, then choose the technology.
Let's build
Tell us the constraint that's holding things up.
Ready to bring AI in-house, scale your operations, streamline a process, or build your next core product? We'll come back with an architecture and a timeline — not a brochure.