Enterprise Products
We build our own ERPs — like CORA, taking on SAP.
Kalnyx
We don't prompt models. We engineer intelligence.
AI-native software, shipped.
Enterprise ERPs · Custom apps · A fully offline AI film studio.
What we build
Three pillars, one engineering core.
Our own enterprise products, custom client software, and an offline AI film studio.
Most companies do one of these. We run all three from one engineering core: CORA taking on SAP, client platforms like the GSV site, and a private generative studio that runs entirely on-prem.
Custom Software
Web & mobile apps built for clients, AI-native.
AI Film Studio
A completely offline AI filmmaking pipeline.
Why us
Most companies rent intelligence. We engineer the whole machine.
- Wire up someone else's API and hope it holds.Own the model, the pipeline, and the product end to end.
- One layer — front-end, or ML, or design.One core that ships the ERP, the app, and the film studio.
- Your data leaves for a cloud you don't control.A generative studio that runs fully offline, on our own metal.
- Hand off between agency, vendor, and freelancer.One pair of hands and a network of agents, start to finish.
That's not a wider service menu. It's a shorter path from idea to a system you own.
Flagship · CORA
The SAP-class construction ERP — one codebase, owned end to end.
A multi-tenant ERP for construction and real estate — procurement, RA bills, vendor payments, materials, geo-fenced labour attendance, GST and Indian labour compliance — with row-level tenant isolation and ML that forecasts delays and cost overruns. One codebase, running on PC, tablet and mobile.
- ProcurementPurchase orders, RA bills & vendor payments.
- MaterialsHierarchical stock, wastage & GST-aware costing.
- AttendanceGeo-fenced smart attendance & labour muster.
- ComplianceGST returns & Indian labour-law compliance.
- IntelligenceML delay, overrun & concrete-strength forecasts.
- AccessMulti-tenant RLS, 7 roles, audit-ready.
How it runs
The engine — and the network that runs it.
Real pipelines and agents, built and run by a tight senior crew.
Next.js and React on the front; an offline generative stack for video, voice and 3D — orchestrated by our own agents. Built by a small team spanning product, AI/ML, design, and 3D.
- Next.jsServer-rendered product front-ends.
- ReactComponent architecture at scale.
- Generative MediaIn-house image & video generation.
- Video GenIn-house AI video pipeline.
- VoiceConsented voice synthesis & cloning.
- 3D PipelineProgrammatic 3D asset pipeline.
- OrchestrationAgent orchestration across tools.
Tight, senior, and operating at the frontier.
- Product engineering
- AI / ML pipelines
- Design & motion
- 3D / generative media
Live build
Watch the systems get written.
The same stack, typed in real time — CORA, the offline studio, the agent layer.
No slideware. These are slices of the systems we actually ship: an ERP core, an on-prem render pipeline, and the agents that drive them.
○ READY
# CORA — FastAPI + SQLAlchemy 2, multi-tenant with RLS@router.post('/ra-bills/{bill_id}/certify')async def certify(bill_id: int, user: User = Depends(require_role)): async with tenant_scope(user.tenant_id) as db: bill = await db.get(RaBill, bill_id) if bill.created_by == user.id: raise Forbidden('maker cannot certify own bill') bill.status = 'CERTIFIED' await audit.log(user, 'ra_bill.certify', bill_id) return bill$ POST /api/ra-bills/{id}/certify
▸ certify(bill=RA-4821) tenant=acme
✓ RLS tenant guard set — row-level isolation on
✓ segregation-of-duties: certifier ≠ maker
✓ RA-4821 CERTIFIED → routed to approver
The network
One founder, a network of agents.
The team behind Kalnyx isn't a payroll — it's a network of AI agents, each owning a slice of the stack. The founder directs; the agents build, render, and ship.
Build Agentshipping Product engineering
- Next.js
- React
- TypeScript
Pipeline Agentrendering AI / ML pipelines
- Video
- Voice
- Image
Design Agentcomposing Design & motion
- UI / UX
- Motion
- Brand
Media Agentgenerating 3D & generative media
- 3D
- Motion
- Assets
Orchestratorcoordinating Agent conductor
- Planning
- Routing
- Tools
The human
Built by Anish Gardas.
The path from first builds to a real company.
FounderHow we deliver
From idea to launch in 12 weeks.
A clear, weekly cadence — discover, design, build, harden, launch.
Deep dive into the problem, users, and current systems.
- Scope & JTBD
- System audit
- Success metrics
Architecture, UX, and a clickable prototype to lock direction.
- Architecture
- UX + UI
- Prototype
Ship the system test-first, integrating in weekly increments.
- Core features
- Integrations
- Weekly demos
Security, performance, and QA until it is production-grade.
- Security pass
- Load + QA
- Polish
Deploy, hand over, and stand up monitoring.
- Deploy
- Handover
- Observability
Trust · on-prem
Nothing leaves the building.
Our generative studio runs entirely on-prem — models, data, and GPUs on our own metal. No third-party API ever sees your footage, your voices, or your IP.
On-prem perimeter
- Models
- Data
- GPUs
- Renders
- Voices
Public cloud APIsshut out
- On-prem modelsEvery model runs locally. No cloud inference, ever.
- Your IP stays yoursFootage, voices, and prompts never leave the perimeter.
- Consented voice onlyVoice synthesis on explicit, logged consent.
- Air-gappableThe studio can run with the network cable pulled.