Web & Digital Products
Corporate websites, platforms and web systems built with performance and visual storytelling.
The full script, from plan to launch.
Mobile Apps
Native and hybrid apps for iOS and Android, with a smooth experience from the first scene to the last.
Cloud & Infrastructure
Cloud architecture, CI/CD and observability: the backstage that keeps the production on air.
Process Automation
Integrations and automations that cut rework and speed up the pace of your operation.
Artificial Intelligence
AI solutions applied to your business's real data, with no needless special effects. See LLMaaS, our private AI platform, below.
Technical Consulting
On-demand technical direction to unblock decisions and speed up critical deliveries.
Private artificial intelligence, running on GPUs in Brazil.
NGG Tech's LLMaaS delivers language models on our own servers, with an OpenAI-compatible API, your company's knowledge base and every token metered by user and by key. A fixed monthly fee in Brazilian reais, taxes included.
- NVIDIA GPUs serving the models with vLLM
- Under 1 minute from clicking activate to a tested environment
- Data in Brazil, under the LGPD
The platform, piece by piece
Six pieces that work together. Pick one to see how it works, or let the panel show them all in sequence.
From contract to first token in under a minute
No waiting on the IT team. Once the purchase is activated, our deployment system sets up your environment on its own and only reports it ready after testing it.
- 1
You pick the plan
In the portal, with a Brazilian CPF or CNPJ, and accept the Terms.
- 2
The purchase is activated
Your company gets its own space, with the plan's quota and limits.
- 3
The environment is deployed
The system checks 10 prerequisites and enables the plan's models.
- 4
Everything is tested
A temporary key makes a real API call and is revoked at the end.
- 5
You start using it
Through the API, in the portal chat or in your own systems.
In the first real deployment, the environment was ready and tested 58 seconds after the activate click.
Every token metered, whatever the path
In-house model, knowledge base or external model: usage goes through the same gateway and shows up on the same dashboard, split by company, user and key.
| Feature | Shared models | Knowledge base (RAG) | External models (your key) |
|---|---|---|---|
| What it is | Qwen3-VL 8B and Llama 3.1 8B on our GPU | Answers based on your company's documents | Claude, GPT and Gemini, with your provider's key |
| What counts | Input and output tokens of each call | Reading each document at ingestion, and the queries | Usage is logged on the dashboard, by key and user |
| Who bills | Included in the plan's monthly allowance | Included in the plan's monthly allowance | Your provider, straight to your account |
| Controls | Requests-per-minute limit, monthly quota, an alert when usage passes 80%, extra tokens when you need them, per-person keys with an expiry date. | ||
Security that comes built in
Your data never leaves our servers in Brazil and trains no model. The protections below apply to every plan.
Per-company isolation
Each company has its own space in the gateway and its own document base. One company never reads another's base.
Keys that expire
Per-person API keys, valid for up to 90 days, with instant revocation and acceptance of the Terms on record.
Approved IPs per key
Your company's administrator decides where each key can be used from.
Secrets vault
Every password and key in our infrastructure is stored in a vault the moment it is created. No secret lives in code or in a spreadsheet.
Closed network
Only the public API is exposed, with TLS encryption. Everything else sits behind a firewall, with no internet access.
Access log
Access is logged for 6 months, and your usage can be audited by user and by key.
Change the URL and go
The API follows the OpenAI format, so the libraries and tools your team already uses work with no code changes.
curl https://api.llmaas.com.br/v1/chat/completions \
-H "Authorization: Bearer $LLMAAS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3-vl-8b",
"messages": [
{"role": "user", "content": "Summarize the contract in 5 points."}
]
}'/v1/chat/completionstalks to the models/v1/completionscompletes text/v1/modelslists the models in your plan/v1/knowledgesends documents to your company's base, with the same key
Plans with a fixed monthly fee
No surprise per-token charges: every plan comes with a monthly allowance.
| Feature | Starter | Pro | Enterprise | Custom |
|---|---|---|---|---|
| Tokens per month | 5 million | 50 million | 500 million | tailored |
| Requests per minute | 60 | 300 | 3,000 | tailored |
| Storage | 5 GB | 50 GB | 500 GB | tailored |
| In-house GPU models | Qwen3-VL 8B | Qwen3-VL 8B and Llama 3.1 8B | the full catalog | your choice |
| External models with your key | — | Claude, GPT, Gemini | Claude, GPT, Gemini | your choice |
| Knowledge base (RAG) | — | — | yes | your choice |
| Dedicated inference | — | — | on request | tailored |
| Support | community | priority e-mail | dedicated 24/7 | tailored |
External models are billed by your provider.
Frequently asked questions
Does my data train any model?
No. Calls and documents stay on our servers in Brazil and are not used to train models.
Do I need to change my code?
Hardly at all: the API follows the OpenAI format. Just change the URL to https://api.llmaas.com.br/v1 and use the key created in the portal.
What happens if the monthly quota runs out?
You get an alert when usage passes 80%. If you need more, buy extra tokens or change plans in the portal.
How does using Claude, GPT and Gemini work?
You store your provider's key in the portal. Calls go through our gateway, which logs the usage, and the provider bills your account directly.
Which documents does the knowledge base accept?
PDF with selectable text, Word (.docx), TXT, Markdown and HTML, up to 50 MB each, plus web pages. Answers cite the file and page they came from.
How long does it take to get started?
Once the purchase is activated, the environment is ready and tested in under a minute.
Shall we talk about your project?
Tell us what your company needs, from a website to an AI platform, and we'll write the script with you.