bampa AI · on premises

Large models.
In your server room.

We design and install AI servers that run top-tier open-source models for the whole company, on your documents, your code and your projects. No pay-per-use subscriptions, no data leaving the building.

bampa server · on premises EPYC 8 channels · ~205 GB/s GPU GPU + GPU when you need it
bytes of data leaving the company
0
people served at once by a single server
60
correct answers with the source, guaranteed in the field trial
16/20
always available, no pay-per-use costs
24/7

Use cases

What you can do with it,
starting Monday morning.

bampa · internal server internet: disconnected
What's the tightening torque for the spindle screws on the P200?
reading 3 internal documents…

For the P200 the specified torque is 38 N·m, crosswise in two passes: first at 50%, then at full value. Since revision 4 this also applies to spindles with an HSK interface.

Manual_P200_rev4.pdf · p. 112 PQ-07 Torques.docx · §3.2

Illustrative example: every answer cites the document and the page.

  • Company know-how

    Manuals, procedures, specifications and standards you can query in plain language, with the source cited.

  • Software development

    An assistant on your private code: reviews, tests, documentation, legacy code to explain and migrate.

  • CAD design

    The model drives CAD through deterministic tools: parametric parts, assemblies, bills of materials, dimensional checks.

  • Electronics and PCBs

    Schematics, netlists, footprints, routing rules, reading DRC and ERC errors, normalised BOMs.

  • Documents and data

    Summaries, version comparisons, extraction of tables, clauses and parameters from thousands of PDFs.

  • Tenders and quotes

    Reading tenders, checking requirements, a first draft of the quote on your own templates.

  • Quality and procedures

    Work instructions always at hand, consistency checks across documents, audit support.

  • Training and support

    The expert colleague available to new hires and internal help desks, around the clock.

  • Meetings and translation

    Transcripts, minutes and technical translation, without recordings shipped to outside servers.

  • Agents and automation

    Workflows that read, check and report on their own, with permissions and a log of every action.

Why local

Your know-how
works for you, and stays yours.

The cloud is convenient until it comes to manuals, drawings, code, quotes and contracts. For subcontractors, medical, defence or anyone under NDA, that's where the conversation ends. On premises, it starts again.

  • 01

    Total confidentiality

    Documents, conversations and code never leave the company network. It works with the internet cable unplugged.

  • 02

    Know-how stays in-house

    Forty years of procedures, solutions and tricks of the trade become searchable, even after the people who know them retire.

  • 03

    Predictable costs

    No charge per generated word: the hardware is depreciated, and the more you use it the better it pays.

  • 04

    No vendor lock-in

    Interchangeable open-source models: when a better one comes out, you switch with one line.

  • 05

    The whole company at once

    Dozens of people at the same time, with per-department permissions and a log of who saw what.

  • 06

    Simpler compliance

    GDPR, trade secrets, customer clauses: if the data never leaves, half the questions disappear.

  • 07

    Built into your tools

    ERP, archive, code editors, CAD and EDA, through standard interfaces (API and MCP).

  • 08

    Always available

    No usage caps, no queues, no service changing its terms overnight.

  • 09

    Tailored to you

    The model learns your vocabulary, your abbreviations and the way you write procedures.

The architecture

EPYC plus several GPUs:
the best of both worlds.

A large model isn't sized by the power of the graphics card, but by memory: how much there is and how fast it is. That's why we combine GPU memory with the RAM of an 8-channel server processor.

GPU

Fast video memory

This holds the most-used layers of the model and the users' context. More GPUs mean more memory and more people served at once, each with their own space.

EPYC · 8 channels

Large server RAM

Hundreds of GB of error-correcting memory for models that don't fit in the GPUs, long contexts, document indexes and always-on agents.

Platform

Grows with you

RAM, GPUs and storage are added one at a time. Start with what you need today and expand when usage calls for it.

Try sizing it

How much server do you need?

Pick the model, how many people use it at once and how much text they work on each time. The figure is indicative: real sizing is done on your own cases.

The model

Memory needed

0 GB

model 0 GB · user context 0 GB

Indicative configuration

—

VRAM
—
RAM
—

Ask for a real sizing

The two-model workflow

The cloud designs the method.
The local model does the work.

When you need the most capable model on the market, use it only to invent the procedure, on a non-confidential example. Then the local model applies it to hundreds of real files that never leave.

  1. Cloud

    Discovery

    understands the problem on a clean example

  2. Cloud

    Formalisation

    writes rules, tools, examples and tests

  3. Local

    Execution

    applies the procedure to all your data

  4. Local

    Verification

    checks format, consistency and errors

  5. Cloud, if needed

    Hard cases

    only the few ambiguous cases, anonymised

How to start

Small, and measured.

  1. 1

    Assessment

    One or two days on site: documents, use cases, sizing, comparison with the cloud. You get a document, not a promise.

    from €1,500
  2. 2

    Field trial

    Two weeks on real data with 5–10 users. At least 16 correct answers out of 20, with the source: otherwise we refund half.

    from €3,500
  3. 3

    Server and installation

    Sized hardware, models, archive, permissions, backup, two-level training.

    on quotation
  4. 4

    Support

    New models as they come out, new documents, monitoring. The system improves while you use it.

    monthly fee

Indicative prices, excluding VAT. Technical documentation and reference architecture on request, under a non-disclosure agreement.

Contact

Let's talk.

A robot, an AI server for your company or a project that doesn't have a name yet: drop us a line and we'll reply within one working day.

info@bampa.net

Based in Auronzo di Cadore and Bolzano, Italy.

The form opens your email app: your data never passes through third-party services. Privacy

Interested in