Skip to content
Cerebra BioSystems

The steps

Eight steps from first call to everyday use.

Durations are typical. Hardware delivery depends on the vendor, and IT prep usually happens while it ships.

  1. Step 1: Free audit30 min
  2. Step 2: Pilot on your data4–6 weeks
  3. Step 3: Hardware spec & ordering1–4 weeks
  4. Step 4: IT prep1 week
  5. Step 5: Install & configure2–4 days
  6. Step 6: Validate3–5 days
  7. Step 7: Handover & training1 day
  8. Step 8: Ongoing supportMonthly

Steps 5–7: Typically 1–2 weeks from hardware arrival to a validated system

Step by step

What happens, what we need, what you get.

  1. Step 1: Free audit

    30-minute call · recommendation within a few days

    What happens
    We talk through your instruments, run volumes, assays, current analysis setup, IT requirements and goals. If it helps, you show us the size and layout of a typical run folder, so storage is sized from real numbers.
    What we need from you
    • 30 minutes with the lab director, and your IT lead if possible
    • Rough run volumes and the assays you run
    • Any IT or security rules we should know about early
    What you get

    Written recommendation

    One page: what to automate, the hardware to buy and an estimate.

  2. Step 2: Pilot on your data

    4–6 weeks · fixed fee

    What happens
    One assay, your real data, at a fixed fee. We re-analyse recent runs and compare them with your existing results, and your team uses the private AI on them. It runs in your lab, on hardware we agree during the audit.
    What we need from you
    • Two or three recent runs, with the results you already have for them
    • One or two scientists to try the private AI and tell us what's missing
    • Success criteria agreed up front, for example results and a summary within 24 hours of a run finishing
    What you get

    Pilot report

    The side-by-side comparison with your results, what worked, what didn't, and a recommendation on whether to go ahead.

  3. Step 3: Hardware spec & ordering

    Set by vendor delivery, often 1–4 weeks

    What happens
    We write a spec sized to your run volumes, assays and how long you keep data. You order directly from the vendor at their price; we don't resell or mark it up. It ships to your lab, or to us first so we can set it up and test it before it arrives.
    What we need from you
    • Purchase approval and an order placed with the vendor
    • A delivery address and someone to receive it
    • Any preferred vendors your procurement team uses
    What you get

    Hardware spec sheet

    Models or equivalents, quantities, storage sizing with its assumptions, and power and network requirements.

  4. Step 4: IT prep

    About a week, while the hardware ships

    What happens
    Your IT lead and we agree where the hardware sits, its network addresses, firewall rules, company sign-in and where backups go. It's usually one call and a few emails, working through the IT checklist below.
    What we need from you
    • Answers to the IT checklist
    • A named admin on your side
    • An app registration in your identity provider, for sign-in
    What you get

    Network and access plan

    Addresses, VLAN, firewall rules, sign-in setup, backup target and who holds admin rights, on one page.

  5. Step 5: Install & configure

    2–4 days, mostly on site

    What happens
    If the hardware comes to us first, we install the software, load the AI model and reference genomes, and run our test suite before it ships on. On site, we place it on a desk or in a rack, connect it to your network and sequencer storage, set up company sign-in and encrypted connections, and configure your pipelines.
    What we need from you
    • Access to the room and the network on the agreed days
    • Your IT lead for a few hours, for network and sign-in changes
    • A gap between sequencing runs to connect the output share
    What you get

    Install record

    What was installed and where, the software and pipeline versions, and how everything is configured.

  6. Step 6: Validate

    3–5 days

    What happens
    We run reference datasets with known results, re-analyse one or two of your past runs and compare them with your previous results, check that repeat runs agree, and test the AI against questions with known answers.
    What we need from you
    • One or two past runs, with the results you trust
    • A scientist to review the comparison with us
    What you get

    Validation report

    What we tested, the comparisons, any differences and why, and a sign-off page.

  7. Step 7: Handover & training

    1 day, then 2 weeks of hypercare

    What happens
    A one-hour session for your scientists on the private AI and run reports, and an admin handover for IT. Then two weeks of hypercare: we check in often and fix anything that comes up.
    What we need from you
    • An hour with the scientists who'll use it
    • Your admin, for the handover session
    What you get

    Admin guide and runbook

    How to add users, switch remote access on and off, check backups, and what to do when something fails.

  8. Step 8: Ongoing support

    Flat monthly fee

    What happens
    Monitoring if you enable it, security updates, tested model and pipeline updates you approve, fixes, a monthly allowance of analysis help and quarterly reviews.
    What we need from you
    • Approval for updates and maintenance windows
    • Remote access switched on when we need it, or an on-site visit booked
    What you get

    Quarterly review

    Runs processed, usage, issues, updates and what to automate next.

Hardware we specify

Hardware you own, sized to your lab.

You buy it directly at vendor pricing; we specify, install and maintain it.

Hardware illustration: your sequencer writes run folders to network storage, a compact AI computer works on those runs and serves the private AI workspace to your team's browsers, and an optional pipeline server takes on the pipelines in larger setups. Everything sits inside your network.Your networkOn a dedicated port or VLANYour sequencerWrites run foldersrun foldersNetwork storageRuns, results, snapshotsPipeline serverOptional:Standard, Scaleruns and resultsCompact AI computerPrivate AI + pipelinesAI workspaceYour teamBrowser on your network
Starter setupAdded in Standard and Scale
Text description of the hardware

Everything sits inside your network, on a dedicated port or VLAN. Your sequencer writes its run folders to network storage, which holds raw runs, results and snapshots.

In the Starter setup, a compact AI computer reads the runs from storage, runs the pipelines, writes the results back, and serves the private AI workspace. Your team opens the workspace in a browser on your network.

In the Standard and Scale setups, a separate pipeline server (drawn dashed) runs the pipelines against the same storage, and the AI computer focuses on the private AI.

Typical builds

  • Starter

    A compact AI computer plus network storage.

    Typical fit
    Up to about four sequencing runs a month, mostly bulk RNA-seq or amplicon, with a small team.
    What's in it
    • One compact AI computer that runs both the pipelines and the private AI
    • Network storage with 4–8 drive bays, about 40–80 TB usable
    • A fast network switch and a battery backup (UPS)
  • Standard

    An AI computer, a dedicated pipeline server and larger storage with backup.

    Typical fit
    Weekly runs, several assays, or whole-genome work.
    What's in it
    • A compact AI computer for the private AI
    • A dedicated pipeline server
    • Network storage of about 100–200 TB
    • A second storage target for backups, and a UPS
  • Scale

    Multiple compute nodes and enterprise storage.

    Typical fit
    High-volume sequencing, or a CRO serving many sponsors.
    What's in it
    • Two linked AI computers, or one GPU workstation
    • Several pipeline servers working together
    • Enterprise storage sized to your sponsors and retention

You buy it directly at vendor pricing. We specify, install and maintain it, and we don't resell or mark it up. Sizes are typical: we size yours from your real run folders during the audit.

Estimate your setup

IT prerequisites

What your IT lead needs to have ready.

Most of this is settled in one call while the hardware ships. If something here doesn't fit how you work, tell us and we'll plan around it.

  • Power and space

    A desk, shelf or rack spot near a network port, with standard outlets. The spec sheet lists power draw, and includes a battery backup (UPS) if you don't already have one.

  • A network port or VLAN

    A wired connection (10 Gb if you have it), a dedicated VLAN if your policy calls for one, and a few static IP addresses.

  • Access to the sequencer output share

    Read access to the folder your sequencer writes runs to, or agreement to point the sequencer at the new network storage.

  • Firewall rules

    Connected: outbound-only rules for approved updates, optional health metrics and remote support when you switch it on. No inbound ports. Air-gapped: no internet rules at all.

  • Sign-in (SSO) details

    Your identity provider (Microsoft Entra ID, Google Workspace or another OpenID Connect provider), an app registration for the workspace, and the groups that map to projects.

  • Who the admin is

    The person who approves updates, manages users and switches remote access on or off. It can be your IT lead, a lab manager or your outside IT provider.

  • A backup target

    Where the second copy lives: a second storage box, your existing backup system, or offline drives you rotate.

it-checklist.txt
Cerebra BioSystems · IT checklist

[ ] Space     Desk, shelf or rack spot near a network port
              Standard outlets; UPS included if you need one
[ ] Network   Wired port (10 Gb if available) or dedicated VLAN
              A few static IP addresses
[ ] Storage   Read access to the sequencer output share
[ ] Firewall  Connected: outbound only, no inbound ports
              Air-gapped: no internet access at all
[ ] Sign-in   Identity provider (OpenID Connect)
              App registration + groups mapped to projects
[ ] Admin     Named person for updates, users, remote access
[ ] Backup    Second storage box, backup system or offline drives
Copy it into a ticket or an email. The full version comes with your hardware spec sheet.

Validation & acceptance

Tested against known answers before your team relies on it.

You accept the system against criteria we agree at the start, using results you can check yourself.

  1. Datasets with known results

    Public reference datasets with published results go through your pipelines first. Then we re-analyse one or two of your past runs, compare them with the results you already trust, and explain any differences.

  2. Reproducibility checks

    We run the same input twice and compare the results. Pipeline versions, container images and input checksums go into a run record, so any run can be repeated later.

  3. AI answer test set

    Questions with known answers about your validation runs. We check that each answer is right, that every number links to the file and row it came from, and that the AI says “not found” when the data isn't there.

  4. Sign-off

    You review the validation report against the success criteria agreed at the start, and sign off when it meets them. The same tests run again before every update.

Validation report

Example outline
  1. 01Scope: pipelines, versions and reference genomes
  2. 02Reference datasets: expected and observed results
  3. 03Your past runs: comparison with previous results
  4. 04Reproducibility: repeat-run comparison and checksums
  5. 05AI test set: correct answers, sources, “not found” cases
  6. 06Differences found, and their causes
  7. 07Acceptance against your success criteria

Your lab director

Signed and dated

Cerebra BioSystems

Signed and dated

After go-live

Kept running, watched and up to date.

Two weeks of hypercare, then flat monthly support. Nothing changes on your system without your approval.

  • Continuous, if enabled

    Monitoring

    Hardware and pipeline health from the metrics you choose to share. Never data, file names or conversations.

  • When released

    Validated updates you approve

    New model and pipeline versions are tested on your system with the validation tests before cut-over. You approve each one, and the previous version stays available for rollback.

  • As needed

    Pipeline maintenance

    Fixes when a run fails, reference genome updates, and changes when your assays or instruments change.

  • Every month

    Analysis help

    A monthly allowance of help from a bioinformatician, for questions beyond the standard reports.

  • Every quarter

    Quarterly review

    Usage, runs processed, issues, updates and what to automate next.

FAQ

Questions about installation.

Do we need IT staff?
Not full-time. During setup we need someone who can make network and sign-in changes, which can be an outside IT provider. Afterwards, a named admin approves updates and manages user access. We look after the system itself.
Does it need internet?
No. In connected mode it makes outbound-only connections for approved updates and, if you enable them, hardware-health metrics. Nothing connects in from outside. If you'd rather it had no internet connection at all, choose the air-gapped option.
What if we already have a server?
Tell us during the audit. If it has the capacity, an existing server or storage system can often run the pipelines or hold the data, and we size the rest around it. The private AI needs a GPU with plenty of memory, so that part is usually new hardware.
How disruptive is installation?
Not very. Most of the preparation happens before we arrive. On site, we need your network and the sequencer output share for a few hours, scheduled between runs. Your sequencer and current analysis keep working throughout.
Can it run air-gapped?
Yes. In the air-gapped option the system has no internet connection at all. Updates arrive on encrypted media and we install them on site, after the same tests as connected updates. Support happens during on-site visits.

Find out what your lab could automate.

30 minutes, no cost. You get a written recommendation afterwards.

30 minutes · No cost · No obligation · We'll send a written recommendation afterwards.