SAM_SSN_Posting_1.pdf

United States Department of the Interior

INTERIOR BUSINESS CENTER

Washington, DC 20240

Computational Infrastructure — GPU Server Node

The Department of the Interior Acquisition Services Directorate (AQD) is issuing this Request for Information to determine businesses capable of providing a computational infrastructure – GPU Server Node that meets the salient characteristics listed within this request. AQD is executing this market research on behalf of the Department of Health and Human Services (HHS), National Institute of Allergy and Infectious Diseases (NIAID). The GPU Server Node is part of Phase 1 of the Autonomous Human Biology Pilot Platform, which is described in Attachment 1. However, information is only being requested for the GPU Server Node.

Questions related to the announcement must be submitted no later than 9:00 AM ED July 20, 2026. Respond to this request no later than 12:00 PM ED July 23, 2026, providing the following information:

1. Company name, Point of Contact Information, System for Award Management (SAM) UEI.

2. Written confirmation that your company can provide the requirements within this announcement.

3. Confirm which NAICS you believe would apply to this acquisition.

4. Identify your company’s size standard in the NAICS you identified.

5. Identify any Authorized Reseller Status or subcontractors that would be involved in providing the requirement (if applicable).

6. Identify any established GWACs or GSA schedules applicable to the scope of this announcement.

7.Market researching pricing (not required).

Respondents should identify their socioeconomic status, including:

• Small Business

• Small Disadvantaged Business (SDB)

• 8(a)

• HUBZone

• Service-Disabled Veteran-Owned Small Business (SDVOSB)

• Woman-Owned Small Business (WOSB)

• Veteran-Owned Small Business (VOSB)

• Other

Responses should be emailed to Melissa_Onyszko@ibc.doi.gov with the email title “Computational Infrastructure — GPU Server Node”

This RFI is being issued solely for market research information and acquisition planning purposes. It does not constitute a Request for Proposal (RFP) or a promise to issue an RFP in the future. This RFI does not commit the Government to a contract for any supply or service whatsoever. Furthermore, the Government is not at this time seeking proposals and will not accept unsolicited proposals. Responders are advised that the Government will not reimburse for any information or administrative costs incurred in response to this RFI; all costs associated with responding to this RFI will be solely at the interested parties’ expense. Not responding to this RFI does not preclude participation in a future RFQ, if one is issued.

Proprietary information and trade secrets, if any, must be clearly marked on all materials. All information received that is marked Proprietary will be handled accordingly. Please be advised that all submissions become Government property and will not be returned. All Government and contractor personnel reviewing RFI responses will have signed non-disclosure agreements and understand their responsibility for proper use and protection from unauthorized disclosure of proprietary information as described in 41 U.S.C. 423. The Government shall not be held liable for any damages incurred if proprietary information is not properly identified.

No telephone calls will be accepted requesting a solicitation package. There is no solicitation currently being issued. In order to protect the procurement integrity of any future procurement that may arise from this announcement, information regarding technical point of contact will not be given and no appointments for presentations will be made.

Salient Characteristics:

System: Computational infrastructure Equipment: 4-GPU server node (NVIDIA RTX 6000), 100 TB storage, 96GB RAM on the GPUs Role: Processes imaging data, trains and runs the AI models, stores the resulting datasets Computational Infrastructure — GPU Server Node Requested configuration: 4× NVIDIA RTX 6000 GPUs, 100 TB storage, single server node.

Function:

• Runs the computer-vision models that analyze brightfield and fluorescence images of stem cell colonies — for example, identifying colony boundaries, estimating growth rate, and flagging signs of differentiation or poor cell health.

• Runs the AI decision-making models that use those image-derived measurements to recommend the next laboratory action (e.g., continue incubation, feed, passage, or flag for review). This is the "closed-loop" component described in the program narrative — the software layer that lets the system plan and adapt without a person in the loop for every cycle.

• Stores and processes the large volume of time-lapse and multi-channel image data generated by continuous, automated imaging of up to 44 plates. This type of high-content imaging data is large and accumulates quickly with repeated timepoints, which is the basis for the 100 TB storage requirement.

• Provides the model-training capacity needed to later extend these same AI models from stem cells to organoids in Stage 2, without requiring a second procurement of compute hardware.

Autonomous Human Biology Pilot Platform

— Phase 1 Prototype Stem Cell (iPSC) Growth, Characterization, and Monitoring System, with Planned Extension to Organoids

1. Purpose and Program Context This procurement supports Phase 1 (FY26) of the Autonomous Human Biology Pilot Platform, a foundational autonomous laboratory effort intended to mature into a transferable capability for the NIH Standardized Organoid Modeling (SOM) Center. The pilot platform integrates laboratory automation and robotics, AI-driven experimental planning, data acquisition, and computational infrastructure into a single closed-loop system, where an AI model can propose an experiment, a robotic system can execute it, the results can be captured automatically, and the AI model can use those results to plan the next step — without requiring a human to manually run each cycle. Hence, an autonomous system.

The equipment requested here is the first concrete build-out of that platform. It is scoped as a multi-stage prototype:

• Stage 1 (this order): Autonomous growth, characterization, and monitoring of induced pluripotent stem cells (iPSCs) — the starting material used to generate organoids. This stage validates the core hardware and software loop on a simpler, well-characterized biological system before applying it to more complex tissue models.

• Later stages (2 & 3) (planned extension, not part of this request): Once the iPSC prototype is operating reliably, the same platform will be extended to organoid culture — including scaling culture conditions and extracellular matrix (ECM) formulations — to generate the additional data needed to extend the AI models from stem cells to organoids.

Framed simply: we are first teaching the system to reliably grow and grade the "seed" (iPSCs) before asking it to grow and grade the more complex "plant" (organoids).

2. What Is Being Procured Two integrated systems are required, one computational and one physical. Neither is useful without the other — the wet-lab hardware generates the data and executes the experiments; the compute hardware turns that data into decisions.

System Equipment Role

Computational infrastructure

4-GPU server node (NVIDIA RTX 6000), 100 TB storage

Processes imaging data, trains and runs the AI models, stores the resulting datasets

Autonomous wet-lab platform

BAB400 robotic cell-culture system with 44-plate incubator, brightfield/fluorescence plate imager, and single-cell/organoid dispense tool

Grows cells under controlled conditions, images them on a schedule, and physically carries out the culture steps (feeding, passaging, dispensing) that the AI recommends

3. Computational Infrastructure — GPU Server Node Requested configuration: 4× NVIDIA RTX 6000 GPUs, 100 TB storage, single server node.

Function:

• Runs the computer-vision models that analyze brightfield and fluorescence images of stem cell colonies — for example, identifying colony boundaries, estimating growth rate, and flagging signs of differentiation or poor cell health.

• Runs the AI decision-making models that use those image-derived measurements to recommend the next laboratory action (e.g., continue incubation, feed, passage, or flag for review). This is the "closed-loop" component described in the program narrative — the software layer that lets the system plan and adapt without a person in the loop for every cycle.

• Stores and processes the large volume of time-lapse and multi-channel image data generated by continuous, automated imaging of up to 44 plates. This type of high-content imaging data is large and accumulates quickly with repeated timepoints, which is the basis for the 100 TB storage requirement.

• Provides the model-training capacity needed to later extend these same AI models from stem cells to organoids in Stage 2, without requiring a second procurement of compute hardware.

Baseline facility requirements: the server will be located in NIAID’s OCICB datacenter, and a wired network connection between the server and the wet-lab instrumentation described below, so that images and instrument data can move to the server automatically and the AI can instruct the laboratory robot regarding next experimental steps.

4. Autonomous Wet-Lab Platform — BAB400 System The BAB400 is a robotic cell-culture and bioprinting platform built around a six-axis robotic arm that swaps between interchangeable tool "hands" to carry out different laboratory tasks on the same set of plates, without a person physically handling them. The specific configuration requested includes three components:

a) 44-plate incubator

• Provides controlled temperature, humidity, and CO₂ environment for continuous, unattended storage and incubation of up to 44 microplates at a time.

• Plates are accessed directly by the robotic arm, so cultures can be maintained around the clock without manual removal from controlled conditions (which is itself a source of variability in standard manual workflows).

b) Brightfield/fluorescence plate imager

• Captures repeated, non-destructive images of the cells in each plate over time (both standard brightfield images and fluorescence images, which can visualize specific labeled markers).

• This imaging is what feeds the AI models on the GPU server — it is the "monitoring" and "characterization" function referenced in the project description, generating the data the models use to assess growth and cell state at each timepoint.

c) Single-cell (organoid) dispense hand

• A robotic tool-head that precisely dispenses individual cells or small cell clusters into plates.

• Used for standardized seeding of new cultures at a consistent density — important because seeding consistency is a major source of variability between plates and between runs in manual workflows. In Stage 2, this same tool is what enables consistent organoid seeding into ECM structures as the platform scales to organoid culture.

Baseline facility requirements: dedicated bench or floor space within a controlled lab environment (per manufacturer footprint and utility specifications), standard lab power, and CO₂ gas supply for the incubator; network connectivity to the GPU server for automated image and data transfer.

5. How the Two Systems Work Together (Concept of Operations)

1. The BAB400 imager captures a scheduled image of each active plate (brightfield and/or fluorescence).

2. Images are transferred automatically to the GPU server.

3. AI models on the server analyze the images to assess colony growth, morphology, and cell state.

4. Based on that analysis, the system determines the next action for each plate — continue incubating, feed, passage, dispense into a new plate, or flag for human review.

5. The BAB400's robotic arm carries out the selected action using the appropriate tool head.

6. All images, measurements, and actions are logged, building the dataset used both to monitor individual cultures and to continue training and improving the AI models.

This loop is what is meant by "closed-loop experimentation" in the program description — each cycle produces the data needed to run the next cycle with less manual intervention, and the accumulated data is what will let the same approach be extended from stem cells to organoids in Stage 2.

6. Relationship to Stage 2 (Not Part of This Order) The prototype above is scoped specifically to stem cell (iPSC) culture. It is being built as a foundation for a planned second stage — not yet procured — that will extend the same hardware and AI models to organoid culture, including testing different scaling conditions and ECM formulations. The compute and imaging capacity requested here is intentionally sized to support that extension, so that Stage 2 can reuse this infrastructure rather than requiring a separate build-out.

7. Summary Requirements for Procurement

Item Quantity/Spec Primary Function

GPU server node 4× NVIDIA RTX 6000 GPUs AI model training/inference on imaging data

Server storage 100 TB Storage of high-content time-lapse imaging datasets

BAB400

platform 1 unit Robotic (6-axis arm) cell culture automation

Incubator module 44-plate capacity Controlled, robot-accessible culture environment

Plate imager Brightfield/fluorescence Non-destructive periodic imaging for AI monitoring

Dispense tool head

Single-cell/organoid dispense

Standardized, reproducible cell/organoid seeding