Skip to content
Cerebra BioSystems

On-premise bioinformatics · Private AI

Your in-house bioinformatics team and private AI — installed in your lab.

Cerebra BioSystems sets up automated sequencing pipelines and a private AI workspace on hardware you own. When a run finishes, results, quality checks and plain-English summaries are waiting for your scientists — and your data never leaves your network.

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

Demo · public RNA-seq dataset

Cerebra · Private AI workspace

Scientist:

Did yesterday's run pass QC?

Read run_0142 / multiqc_general_stats.txt

Cerebra:

23 of 24 samples passed. Sample S07 is flagged: 46% of reads aligned (threshold 80%) and adapter content is high, which usually means adapter contamination during library prep. I'd re-trim and re-run S07 before using it in the differential expression analysis.

  • Source: run_0142/multiqc_general_stats.txt, row S07
  • Source: nf-core/rnaseq, 3.x

Run 0142

Bulk RNA-seq · 24 samples

  1. Demultiplexed — Complete
  2. Aligned & quantified — Complete
  3. QC report — Needs attention1 sample flagged
  4. Differential expression — Complete
  5. Summary ready — Ready to read

nf-core/rnaseq 3.x · pinned

multiqc_general_stats.txt

23 of 24 samples passed.

S07Reads aligned

46%threshold 80%

Built for 5–50 person biotechs and CROs running Illumina, Element, PacBio or Oxford Nanopore sequencers.

  • Bulk RNA-seq
  • Single-cell RNA-seq
  • WGS / WES
  • ATAC-seq
  • Amplicon & 16S

Bulk RNA-seq available now; other assays on request.

Sound familiar?

Sequencing is fast. Getting answers from it isn't.

  • Waiting on one person.

    Your only bioinformatician is the bottleneck for every run, every figure and every “quick question.”

  • Outsourcing that takes weeks.

    Sending FASTQs to a vendor or consultant means waiting, paying per project and shipping your data off-site.

  • AI your legal team won't approve.

    Your scientists want to use AI on their results, but pasting unpublished data into a public chatbot isn't an option.

  • Data that keeps growing.

    Every run adds terabytes, and cloud storage and compute bills grow with it.

How it works

From finished run to answers, automatically.

Four steps: sequencer finishes, pipelines run automatically, results and QC report are produced, scientists ask the private AI questions with cited answers.
  1. Your sequencer finishes.

    Data lands on storage in your lab. Cerebra BioSystems detects the completed run and checks the sample sheet.

    Run folder + sample sheet

  2. Pipelines run on their own.

    Community-validated nf-core pipelines process the data, with pinned versions so every result can be reproduced.

    nf-core · pinned versions

  3. Results and QC, ready to read.

    You get a run report with quality flags, expression results and standard figures. No scripts required.

    Run report · QC flags

  4. Ask your private AI.

    Scientists ask questions in plain English. Every answer shows exactly which file and row it came from.

    Cited answers

Private AI workspace

Ask questions about your data. Get answers you can check.

A ChatGPT-style workspace that runs entirely on your hardware and is connected to your pipeline results, your papers and your protocols.

Cerebra· Private AI workspace

Demo · public dataset
  • QC

    Scientist: Which samples should I worry about in run 0142?

    Cerebra:

    Only S07. Low alignment (46%) and high adapter content. All other samples are within your thresholds.

    • Source: multiqc_general_stats.txt, row S07
  • Expression

    Scientist: Is GENE_X changed in treated vs control?

    Cerebra:

    Yes. GENE_X is up 3.1-fold (log2FC 1.63, adjusted p = 0.0004) in treated samples.

    • Source: de_treated_vs_control.tsv, row GENE_X
  • Failed run

    Scientist: Why did run 0143 fail?

    Cerebra:

    The alignment step ran out of memory on sample S12, whose FASTQ is about 4× larger than the others. Re-running with the high-memory setting will fix it; nothing needs re-sequencing.

    • Source: .nextflow.log, process STAR_ALIGN (S12)

Every number links to its source. No source, no answer.

If the data isn't in your results, the AI says so rather than guessing.

What we install

Hardware you own, sized to your lab.

You buy the hardware directly at vendor pricing. We specify it, install it, and keep everything running.

  • Starter

    A compact AI computer plus network storage. For labs running a few sequencing runs a month, mostly bulk RNA-seq.

    Includes

    • Compact AI computer
    • Network storage
  • Standard

    An AI computer, a dedicated pipeline server and larger storage with backup. For weekly runs, multiple assays or whole-genome work.

    Includes

    • AI computer
    • Pipeline server
    • Storage + backup
  • Scale

    Multiple compute nodes and enterprise storage. For high-volume sequencing and CROs serving many sponsors.

    Includes

    • Multiple compute nodes
    • Enterprise storage

Private AI vs cloud AI

The AI comes to your data — not the other way around.

Cloud AI compared with Cerebra BioSystems private AI
FeatureCloud AICerebra BioSystems
Where your data goesVendor's serversStays on your hardware
Knows your resultsOnly what you paste inConnected to every run
Shows its sourcesRarelyEvery number
Who controls accessThe vendor accountYour company login and permissions
Cost modelPer seat or per useHardware you own + flat support

Security

Designed to pass your IT review.

  • Stays on your network.

    Data, results and AI conversations never leave your building.

  • Your company login.

    Sign in with Microsoft or Google accounts your IT team already manages.

  • You control our access.

    Remote support is off by default. Your admin switches it on when needed, and every session is logged.

  • Full audit trail.

    Every login, question, pipeline run and admin action is recorded.

Working with us

Start small. Prove it on your own data.

  1. Free audit

    30 minutes

    We learn your instruments, runs and goals, then send a one-page written recommendation.

  2. Pilot

    4–6 weeks

    One assay with your real data, at a fixed fee.

  3. Installation

    Hardware set up in your lab, connected to your sequencer and company login, and validated against your previous results.

  4. Ongoing support

    Monitoring, tested updates, new pipelines and analysis help, for a flat monthly fee.

Demo

See it on real data.

A recorded demo is on the way. In the meantime, we're happy to walk you through it live on a public dataset.

Book a live walkthrough

Why Cerebra BioSystems

I've spent my career on the computational side of biology — building pipelines and lab data systems. Small teams shouldn't have to choose between hiring a bioinformatics department and sending their data to someone else's cloud. Cerebra BioSystems gives them a third option.

Filip IsakovicFounder, Cerebra BioSystems

FAQ

Common questions.

Do we have to buy hardware? Roughly what does it cost?
Yes, you own the hardware. We specify it and you buy it directly from the vendor. A starter setup (one compact AI computer plus network storage) is typically in the range of $6,000–$15,000 at current prices, depending on storage size.
Will this replace our bioinformatician?
No. It takes the routine work off their plate: run processing, QC checks, standard reports and “quick questions.” That leaves them time for the analysis that needs their expertise. If you don't have a bioinformatician, it gives you most of what one would set up, plus our support.
Which sequencers and assays do you support?
We work with data from Illumina, Element, PacBio and Oxford Nanopore instruments. Bulk RNA-seq is available now, and single-cell, WGS/WES, ATAC-seq and amplicon pipelines are available on request.
How is this different from BaseSpace or Illumina Connected Analytics?
Those are cloud services: your data is uploaded and processed on Illumina's servers. Cerebra BioSystems runs everything on hardware in your lab, adds downstream analysis and a private AI workspace connected to all your runs, and doesn't charge per sample. Many labs still use their sequencer's built-in processing, and we can pick up from there.
How accurate is the AI? What if it's wrong?
The AI doesn't calculate results. It reads them from your pipeline outputs and shows the exact source for every number, so anyone can check it in one click. If it can't find something, it says so. Before any update, we re-run a test set of questions with known answers.
Does any data leave our building?
No. Data, results and AI conversations stay on your network. In connected mode, the only thing that can leave is optional hardware-health information like disk space and temperature. In air-gapped mode, nothing leaves at all.

Find out what your lab could automate.

In a free 30-minute audit we'll map your instruments, data volumes and analysis needs, then send you a written recommendation, whether or not you work with us.

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