Ask the workspaceGrounded in your documents
Ask anything across your documents…
For principal investigators
Private AI desktop for academic labs and principal investigators.
A dedicated AI environment that helps your lab search prior grants, draft protocols, summarize literature, analyze data, and prepare manuscripts — without exposing unpublished work to outside AI providers.
Book a private AI systems call
For PIs running grant-funded labs with confidential data, unpublished manuscripts, IRB and IACUC protocols, and peer-review obligations.
The problem
You want AI in the lab. You cannot put unpublished work or patient data into ChatGPT.
A lab generates a decade of grant pages, protocols, raw data, manuscript drafts, and peer reviews. Most of it is too sensitive to share — but the work depends on it. Public AI tools are off-limits exactly where they would help most.
Trapped in
- Funded and unfunded grant applications
- Specific Aims drafts across cycles
- IRB and IACUC protocols
- SOPs and lab notebooks
- Raw and analyzed data
- Unpublished manuscript drafts
- Peer review correspondence
- Trainee theses and dissertations
Eaten by manual work
- Re-reading old grants for prior language
- Re-drafting the same Aims paragraph
- Comparing protocol versions by hand
- Pulling methods from scattered SOPs
- Tracking down which dataset a figure came from
- Summarizing literature for an intro section
- Re-finding peer review comments across submissions
- Editing the same grammar mistakes in trainee drafts
Today: scattered, manual, repeatedOne private layer over the same documents
A private AI desktop lets the lab move faster on writing, drafting, and review without unpublished work leaving the institution.
What you build
A private AI layer over your lab’s actual archive.
Trained on your prior grants, protocols, SOPs, and manuscript drafts. Reads the literature you point it at. Cites everything back to a source you can verify.
01
Search across your full lab archive
Plain-English questions over grants, protocols, and drafts.
02
Specific Aims and grant drafting
Pulls forward language that scored well in prior cycles.
03
Protocol drafting and comparison
IRB, IACUC, and SOP versioning side-by-side.
04
Methods section drafting
Composed from the protocol and lab records, with citations.
05
Literature triage
Read, summarize, and tag against your own paper claims.
06
Data Q&A grounded in your own records
Natural-language questions over your spreadsheets and exports.
07
Peer review preparation
Comments organized, line-by-line responses drafted.
08
Trainee writing review
A consistent first pass on theses, drafts, and code comments.
Unpublished work stays inside your institution. The model only sees what you decide to expose.
Principles
How we work, in four lines.
These constraints are baked into every lab deployment. Unpublished work is the asset — the system is designed to keep it inside the institution.
Unpublished work never leaves the perimeter.
Models, documents, and chat run on hardware in the lab or in your institution’s private cloud. Drafts and data are not used to train anyone else’s model.
Every claim cites a real document.
No fabricated references. Outputs point to the page, the file, the version. Where the model has no source, it says so. The behavior is testable on day one.
Per-protocol access scoping.
Trainees see what they need. Patient-linked data is gated by IRB protocol. The PI’s strategy documents stay PI-only — even on a shared system.
Designed around your institution’s policies.
We engage your IRB, IT, and compliance teams before any data is ingested. The system sits inside the perimeter they already trust.
Example deployment
Lab workspace deployment
A reference design for grant-funded research labs. Sized for a single PI’s portfolio rather than a whole department, with the option to extend across collaborating labs under a shared protocol.
Project sizeTens of thousands
ClientNIH-funded research lab
Use caseGrants, protocols, manuscripts, and data analysis
A dedicated AI environment for the lab. Models, document processing, and chat all run inside the institutional perimeter. Reads from the lab archive, the prior grant cycles, and the protocol library. Writes nothing without explicit PI sign-off.
Document types included
- Funded and unfunded grant applications
- Specific Aims and project narratives
- IRB and IACUC protocols and amendments
- SOPs and lab notebooks
- Raw and analyzed datasets
- Manuscript drafts and resubmissions
- Peer review correspondence
- Trainee theses and dissertations
Core workflows
- Specific Aims and grant drafting
- Protocol drafting and version comparison
- Methods section composition
- Literature triage and tagging
- Data Q&A on lab spreadsheets
- Peer review response drafting
- Trainee writing review
- Renewal and progress report prep
Security model
- On-premise or institutional private cloud
- Compliant with university data policies
- Unpublished work never leaves the perimeter
- Role-based access by lab member
- Audit logs on every prompt and retrieval
- Per-protocol access scoping
- Trainees can read but not modify PI documents
- Admin kill-switch
Have a similar workflow?
Book a call →
Reference deployment · NIH-funded research lab
Who this is for
Built for PIs who run their own grants and protocols.
This is for PIs whose competitive edge depends on the archive they have built — prior grants, protocols, datasets, and drafts. Where unpublished work is the asset.
Strong fits
- NIH-funded research labs
- NSF and DOE-funded labs
- Independent clinical research groups
- Translational medicine labs
- Computational biology and bioinformatics groups
- Engineering and physical sciences labs
- Multi-PI program project teams
- Foundation-funded research programs
Common patterns
- Year-round grant deadlines
- Heavy IRB or IACUC administrative load
- Trainees who need a consistent writing first pass
- Datasets too sensitive for public AI tools
- A backlog of manuscripts the PI cannot get to
- Renewals that depend on prior cycle language
- A lab archive too big for the PI to hold in their head
- Reviewer comments that recur cycle after cycle
Especially good fit if you hear
- “I cannot put an unpublished draft into ChatGPT.”
- “Every grant cycle I re-write the same Aims paragraph.”
- “My trainees need a first pass before I see their drafts.”
- “The methods section is the same fight every paper.”
- “I have a decade of grants and no way to search them.”
- “The data is on a shared drive nobody can navigate.”
In their words
What PIs and lab managers tell us.
“I write the same Aims paragraph every cycle. The system pulled my funded language forward without my having to dig for it. The renewal is in better shape than the last submission.”
“My trainees needed a writing first pass. Now they get one — in the lab’s voice, grounded in our SOPs. I see better drafts and we get to mechanism faster.”
“I would never put an unpublished manuscript into ChatGPT. This is the only setup IT and the IRB both signed off on.”
Deployment
Local or institutional-private deployment.
We design around your institution’s data policies, not around a vendor’s product roadmap. The workspace sits inside the perimeter your IRB and IT already trust.
Inputs
Funded and unfunded grant applicationsSpecific Aims drafts across cyclesIRB and IACUC protocolsSOPs and lab notebooksRaw and analyzed data
Your perimeterNothing leaves your network
Private AI desktopLocal models · grounded retrieval · approval gate
Document indexLocal modelsPermissionsAudit log
Outputs
Search across your full lab archiveSpecific Aims and grant draftingProtocol drafting and comparisonMethods section draftingLiterature triage
On-premise rig
A dedicated AI workstation in the lab or department. Models run locally.
Institutional private cloud
For labs whose IT prefers a managed environment. Single tenant, no shared inference.
Open-source models
Llama, Qwen, Mistral, and Whisper run locally. No third-party API calls on the inference path.
Per-protocol access scoping
Trainees see what they need. Patient-linked data is gated by IRB protocol.
Lab-archive ingestion
Years of grants, protocols, and drafts indexed for search and retrieval.
Citation-grounded outputs
Every claim ties back to a document and a page. Nothing fabricated.
The point is to give the lab serious AI capability without turning unpublished work into someone else’s training data.
What happens on the call
Book a 30-minute private AI systems call.
We talk through the lab’s actual work — grant cycle, protocol load, manuscript pipeline, and where the PI’s time is currently going. No AI sales pitch.
We will map
- Where your time is currently going
- Which work is publishable and which is sensitive
- What the lab archive looks like today
- Which workflows AI would help with first
- How this interacts with your institution’s IT and IRB
- Whether a private AI desktop makes financial sense for your lab
Book a private AI systems call
You leave with a rough architecture, a priority workflow map, and an honest view of whether this is worth building for your lab.
Project packages
Three ways to engage.
Most labs either start with an audit or jump to a foundational build keyed to the next grant cycle.
Workflow audit
For PIs unsure where AI helps
We spend a few weeks mapping how the lab actually runs. You leave with a buildable architecture, a priority list, and an honest go/no-go.
Foundational lab build
For PIs ready to deploy before the next cycle
A working AI desktop — lab archive search, grant drafting assistant, protocol comparator, and methods drafting. Tuned to the writing patterns you already use.
Full build + automation
For multi-phase deployments
Foundational build plus literature triage, data Q&A, peer review preparation, trainee writing review, and ongoing support across grant and manuscript cycles.
Lab workspace builds commonly land in the tens of thousands depending on archive size, institutional requirements, and the scope of integrations.
Security
Built for unpublished work.
The lab archive is the asset. The system is designed so that unpublished grants, protocols, data, and drafts never leave your institutional perimeter.
01On-premise or institutional private cloud deployment
02No third-party AI APIs on the inference path
03Unpublished work never sent to a public model
04Per-protocol and per-document access scoping
05Trainees can read but not modify PI documents
06Audit logs on every prompt and retrieval
07Compliant with university data and IRB policies
08Encrypted at rest, encrypted in transit
09Admin kill-switch and retention controls
10Citation-grounded outputs — no fabricated references
A PI who lets confidential work into a public AI tool is one disclosure away from a problem. This is the version that does not put you there.
Who builds this
Founder-led. Designed around your institution.
We engage your IRB, IT, and compliance teams from day one. The same rigor we have brought to industrial private-AI deployments, scaled to fit a single PI’s portfolio. The people who design the system are the people who ship it.
BJ
Brian Joseff
Co-founder
Designs the architecture and the institutional security model. Background in private-cloud AI for regulated environments.
CS
Chandler Supple
Co-founder
Leads delivery with each lab. Background in document- and protocol-heavy operational AI.
FAQ
Common questions from PIs.
Outputs are grounded in the documents you have given the system access to. Citations point to actual pages in actual files in your archive. Where the model does not have a source, it says so rather than inventing one — and that behavior is testable on day one.
Build a private AI desktop for your lab.
For PIs whose competitive edge is the archive they have built — and who cannot risk handing it to a public AI tool.