DPDP Act
Sets duties and penalties for how organisations handle personal data.
AI Factory · Sovereign Super Intelligence (SI)
AI Factory: on-premise LLM serving on your own GPU server
BiltIQ AI Factory is an on-premise LLM platform: the GPU server, open-weight model serving and fleet monitoring that run an organisation's own models. BiltIQ installs it on the buyer's GPU server and runs it as a managed service, so model calls stay on the premises. A subscription on BiltIQ's GPU cluster in India is planned and not yet on sale.
We'll walk you through it on sample data.
- Platform-only site licence, one-time, ex-GST
- ₹35 L
- Managed service per year, ex-GST
- ₹16–18 L
- Site-down first response, business hours IST (Mon–Sat 9:00–19:00)
- 4 h
- AI Factory at an engineering college; setup by BiltIQ, unpaid
- 4 DGX Spark
The problem
You answer for where your SI sends data, and cloud bills are metered
RBI FREE-AI
Recommends a board-approved AI policy.
SEBI AI guidance
The regulated entity answers for model outputs, even from third-party tools.
Cloud models are metered
Every new user adds to the bill.
What it is
The base of the SI platform: GPUs and model serving
AI Factory is the GPU fleet and the model-serving layer. Agent OS sits on top and adds the platform gate and human approvals. Your apps (Agent OS workflows, Business OS, Learning OS) run above both.
Model-agnostic: open-weight models chosen per buyer. NVIDIA Nemotron by default for defence and PSU; Qwen3.8-27B for reasoning where allowed.
- AppsAgent OS workflows (tenders, contracts, claims)Business OS (ERP + CRM)Learning OS (LMS + AI Teacher)
- Agent OS (the core)agents, the platform gate, human approval for consequential actions. 97 agents built; 11 pass all 7 of our quality checks.
- AI Factorymodel serving. Open-weight models, your pick (Nemotron default for defence). Single node per model today; multi-node written, not yet in use.
- GPUyour own GPU server, or our cluster in India (subscription: not yet on sale)
Features
Private LLM for enterprise: what your SI server does
Model serving
Serves the open-weight models behind your SI on your GPUs. One model per node today; multi-node serving is written, not yet in use.
Ops (formerly ATC Ops)
Our fleet manager. It runs our own 4-node GPU fleet, one model per node, and watches the health of your server and services. Monitoring is remote, with your permission and no document access.
Grounding on your documents
Answers come from your own documents, with page citations.
Platform gate (with Agent OS)
Deny by default. In on-prem-required mode, no outbound host can be allowed.
Privacy library
Our MIT-licensed biltiq-privacy library detects Indian personal data such as Aadhaar and PAN and pseudonymises it, set per deployment.
Sizing
We size the server to your workload. Hardware is passed through at cost and is never our revenue.
Fine-tuned domain models (roadmap)
After our next GPU expansion. Not promised before then.
How it works
Deploy, ground, then fine-tune later
Deploy.
We commission the server on site and serve the models. With Agent OS on the site, the platform gate is on from day one.
Ground.
Your SI answers from your own documents, with page citations.
Fine-tune (roadmap).
Domain models come after our next GPU expansion.
Engagement:scoping callpaid pilot (4–6 weeks)site licenceyearly managed service.
Controls
Where your SI runs, and what still touches the cloud
| Part | Own it | Subscribe |
|---|---|---|
| Model inference | Your GPU server | BiltIQ's GPU cluster in India (not yet on sale) |
| Outbound calls | With Agent OS: blocked by default (platform gate). In on-prem-required mode no outbound host can be allowed. Other modes allow an admin allowlist, and we show you the list. | — |
| Monitoring | Remote health checks, with your permission, no document access | Run by BiltIQ |
| Serving | Single node per model today; multi-node written, not yet in use | Same |
| Apps on top | Business OS stores files in cloud storage (GCS or S3) today; on-prem file storage is a roadmap item. The Learning OS LMS keeps course files in AWS cloud storage (India region); AI Teacher embeddings, vision and voice default to cloud services and can be repointed. | — |
Sources: BiltIQ code review, Oct 2026. Patent application 202631094208 (fail-closed data boundary), filed Aug 2026, not granted.
Pricing
GPU server for AI: price, licence and managed service
Own it (ex-GST, ranges only):
- Platform-only site (AI Factory serving, no agent workflows)one-time₹35 L
- Full site licence with Agent OS workflowsone-time₹50–90 L
- Managed service₹16–18 L a year
- Hardwareat cost
- Paid pilot4–6 weeks, credited against the licence₹5–8 L
Indicative hardware, at cost; varies with the market
| Option | Indicative, at cost, ex-GST | Power |
|---|---|---|
| RTX PRO 6000 card, 96 GB | ≈₹12.5–16 L | 600 W peak |
| 4× RTX PRO 6000 server | ₹68–80 L | ≈3.5 kW · ≈1 TR cooling · 6 kVA UPS |
Site power and cooling are yours.
Managed service covers: model and platform updates; remote monitoring (no document access); a quarterly health and usage review; agent tuning. Support: business hours IST (Mon–Sat 9:00–19:00); site-down first response within 4 hours. GPU faults go through the OEM warranty, and we manage the claim.
Subscribe: API tokens at ₹8 per 1M input / ₹28 per 1M output, served from our GPU cluster in India. Subscription: not yet on sale.
Proof
Where our SI platform has been tested
An engineering college
AI Factory installed on the college's own GPUs (4 DGX Spark); setup by BiltIQ, unpaid
NHA PM-JAY AI Challenge 2026 (hackathon)
Won
No paid customer install yet. Videos: AI Factory overview · AI Factory explained.
Key facts
AI Factory key facts
- Platform-only site licence: ₹35 L one-time. Full site licence with Agent OS workflows: ₹50–90 L. Managed service: ₹16–18 L a year. All ex-GST.BiltIQ price list, Oct 2026
- Hardware is passed through at cost: indicative ≈₹12.5–16 L per 96 GB GPU card, ₹68–80 L per 4-GPU server.Indian reseller prices, Sep 2026
- Serving runs a single node per model today; multi-node serving is written, not yet in use.BiltIQ code review, Oct 2026
- Models: open-weight, chosen per buyer; NVIDIA Nemotron by default for defence and PSU; Qwen3.8-27B for reasoning where allowed.BiltIQ code review, Oct 2026
- One install so far: an engineering college's own 4 DGX Spark, set up by BiltIQ, unpaid. No paid customer install yet.BiltIQ records
- Patent application 202631094208 (fail-closed data boundary), filed Aug 2026, not granted.Indian Patent Office filing
- Subscribe tokens: ₹8 per 1M input, ₹28 per 1M output, not yet on sale.BiltIQ price list, Oct 2026
FAQ
Frequently asked questions
What is an on-premise LLM, and can BiltIQ set one up in India?
An on-premise LLM is a large language model that runs on servers you own, inside your building, so prompts and documents never go to an outside model API. BiltIQ installs AI Factory on your GPU server and serves open-weight models on it, so model calls stay on your premises; add Agent OS for the platform gate. BiltIQ runs your SI for you as a managed service from India.
How much does an on-premise AI server cost in India?
Hardware is at cost and varies with the market: about ₹12.5–16 L per 96 GB GPU card, or ₹68–80 L for a 4-GPU server (indicative). The software comes on top: a platform-only site licence of ₹35 L, or ₹50–90 L with Agent OS workflows, plus a managed service of ₹16–18 L a year. All ex-GST.
Which open-weight models run on AI Factory?
AI Factory is model-agnostic: open-weight models chosen per buyer. NVIDIA Nemotron by default for defence and PSU; Qwen3.8-27B for reasoning where allowed. Each model runs on a single node today; multi-node serving is written but not yet in use. Fine-tuned domain models come after BiltIQ's next GPU expansion.
How many users can one GPU card serve?
It depends on your documents, prompt length and model choice, so BiltIQ sizes your SI server to your actual workload during the paid pilot, before you buy any hardware.
Is on-premise AI cheaper than cloud AI?
It depends on your volume. Cloud models are metered, so every new user adds to the bill. Owning your SI costs a one-time licence plus a yearly managed service, with hardware at cost, so the cost per query falls as use grows. At low volume, cloud can be cheaper. BiltIQ works out your case on the scoping call.
Is AI Factory available as a cloud service?
It will be, as the Subscribe tier: API tokens served from BiltIQ's GPU cluster in India, at ₹8 per 1M input and ₹28 per 1M output tokens, ex-GST. The subscription is not yet on sale. Join the 2027 waitlist to hear when it opens. Owning your SI on your own server is on sale now.
AI agents with human approvalAI ERP and CRMAI tutor for collegesAI Factory pricingon-prem vs cloud AI costDGX Spark for LLMs, measuredsovereign AI: own it or subscribe
Size your SI server with an engineer
We'll walk you through it on sample data.