AiServa Platform

Your Compute, Managed in One Console

Pair GPU and CPU servers, group them into pools, run virtual machines and databases, and choose the AI models and storage your team uses.

Get Started Request a Demo

An engineer in a clean server room with racks of GPU servers holding a tablet that shows a dashboard of server tiles

One Server Modal, Every Setting

Tabs for Models, Tools, Database, Vector Database, VMs, Lab Engine and Service.

Servers

Pair a Mac, Windows or Linux machine with one command. Servers run a signed backend and update themselves. Mac and Windows are exercised; Linux is untested. Live

Clusters

Group servers into a compute pool with a scheduler, reservations, an urgent lane and gap filling. Frequencies, scans and conformers split into parts across servers and recombine to the last digit. Live

Lab Engine Tab

Hardware (GPU, memory, driver, CUDA), toolchain, self-test, install or update, and a Last Good version to roll back to. Live

Databases per Server

PostgreSQL, MariaDB or SQLite with logins, backups and query tools. Tested on Mac; Windows and Linux installs are still being tested. Partial

Virtual Machines

Parallels on Mac and Hyper-V on Windows: power, console, SSH, snapshots, CPU and memory, clone and delete. Tested against a simulator only. Partial

Cloud Servers

Create CPU or GPU cloud servers from the console with your organisation's own provider key or the platform's, choose the region per server, set monthly caps, and park idle servers as snapshots. Built and tested against a simulated provider API; first provider Vultr. Partial

AI Models

Local models on a server, Bridge models that run on a server (shown as VYROX AI Combo), or cloud providers. Test, edit, switch on and off. Live

File Storage

A paired server, any S3-compatible store (AWS S3, Cloudflare R2, Backblaze B2, Wasabi, MinIO) or the web server. A resumable Move All Files tool. Live

Operations

See What the Team Is Using

Lab Operations: queue with filters, cancel and run again, plain-language failure diagnosis
30-day GPU against CPU seconds by module, server and person
Token Usage: per-person hourly AI metering, with a CSV export
Releases: fleet view, pause, staged roll-out, drift and rollback

Remote Desktop and Terminal

Not Built Yet

Shared Servers

Platform-owned servers offered to entitled organisations with fair share and metering are a design, not a product. Design Only

Multi-GPU and HPC

One computation across several GPUs, or across servers on a fast LAN, waits for suitable hardware to test. Design Only

Shared Hardware Isolation

Isolating several organisations on one GPU is not built, so each organisation computes on its own servers.

Infrastructure Questions

Does AiMolSim compute in the cloud?

No. The web portal does not compute. Computations run on servers you pair with your organisation: GPU workstations, Macs, Linux or Windows machines. You decide which server handles which work.

What is a cluster?

A pool of your servers with a scheduler, a queue, reservations and an urgent lane. Frequencies, scans and conformer searches are split into parts that run on different servers and recombine identically. Multi-GPU runs and fast-LAN HPC are designed but not built.

Which databases are supported?

PostgreSQL, MariaDB or SQLite can be installed and managed per server from the portal, each with its own login, backups and query tools. Apps choose their database in their settings, and Chemistry Lab study data can live there. Windows and Linux database installs are still being tested.

Can I manage virtual machines?

The portal drives Parallels Desktop on Mac and Hyper-V on Windows: power, console, SSH setup, snapshots, CPU and memory changes, clone and delete. So far it is tested against a simulator of those tools, so we mark it partial until it passes on real hosts.

How are servers updated?

Servers run a signed backend and update themselves. A release can be paused, rolled out in stages, and rolled back, and a bad update restores the last good version after three failed starts.

Run Your Next Molecule Through AiMolSim

Sign in, describe a molecule and the question you have, and approve the plan the assistant proposes. Or request a demo first.