UConn Inst for AI Innovation Tech Sol RFP CDR010826 Q&A Addendum 5.pdf

RFP Title: AI Innovation Institute Technology Solutions

RFP#: CDR010826

RE: Question and Answer Addendum 5 Date: February 11, 2026

The following questions and answers will be considered part of the solicitation documents.

Question 1: Given the potential contract term of up to nine (9) years, will proposals that emphasize architectural portability and workload mobility across on-premises and cloud environments be viewed favorably under the “System Architecture” evaluation criteria?

Answer 1: In the competitive process, the University will not pre-evaluate the proposal.

Question 2: Does the University have any expectations that proposers should address transition, exit, or platform handoff strategies at major contract milestones (e.g., end of initial term or extensions)?

Answer 2: The proposer will articulate the proposed project plan and timeline in Question Group 2.2. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 3: With FISMA High identified as the target level, should proposers assume the Federal Compliance Enclave must be architected from inception to support scrutiny from multiple federal granting agencies, even if initial workloads do not immediately require the full control set?

Answer 3: The proposer will articulate the proposed system architecture, project plan, and timeline in Question Group 2.2. The description of how the design supports the University’s ability to meet FISMA requirements may be articulated in Question 2.2.12. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP. Research that does not require such a high level of security will run in a different enclave.

Question 4: When evaluating managed services for regulated enclaves, will the University place additional value on continuous compliance capabilities (e.g., ongoing control monitoring and evidence collection) versus point in time compliance approaches?

Answer 4: In the competitive process, the University will not pre-evaluate the proposal.

Question 5: Does the University expect proposers to include platform level AI governance capabilities (e.g., model lineage, auditability, bias detection, explainability), or should governance be assumed to occur at the individual project or researcher level?

RFP# CDR010826

Question and Answer Addendum 5

Answer 5: It is left to the discretion of the proposer to include proposed software solutions and models that it believes best fulfills UConn’s needs in the proposal.

Question 6: Should proposers anticipate that institution-wide AI governance or usage policies may be introduced during the contract term and ensure the proposed solution can centrally enforce such policies?

Answer 6: Refer to Answer 5 above.

Question 7: Are there informal expectations regarding time to access for new principal investigators or research groups once approval is granted (e.g., same day, multi day, or longer provisioning timelines)?

Answer 7: Provisioning timeline expectations have not been established. It is left to the discretion of the proposer to address service capabilities and related timelines in the proposal.

Question 8: Should proposers assume a broad range of user sophistication (students through advanced researchers), and would differentiated user experience or access models be considered beneficial?

Answer 8: Refer to Addendum 2, Question and Answer 99. Refer also to RFP Section 2.0. Yes, the resource will serve as foundational infrastructure and address the broader AI needs of the University and the range of its stakeholders. The University will not pre-evaluate the proposal.

Question 9: Beyond infrastructure delivery, does the University view deeper strategic partnerships (e.g., joint research initiatives, centers of excellence, or embedded industry collaboration) as additive within the “Value Added Considerations” evaluation category?

Answer 9: It is left to the discretion of the proposer to articulate value-added considerations etc., in the proposal. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 10: Will formal AI workforce enablement offerings (faculty development, student training, certifications, or continuing education) be considered material value added components rather than ancillary services?

Answer 10: The University will not pre-evaluate the proposal. Refer to Answer 9 above.

Question 11: If optional proposer presentations or demonstrations are conducted, does the University anticipate prioritizing deep technical demonstrations or end to end research lifecycle scenarios spanning general, HIPAA, and federal enclaves?

Answer 11: Refer to Addendum 3, Question and Answer 42.

Question 12: Can UConn provide examples of AI use cases that its researchers are planning to develop and deploy on this new platform?

Answer 12: Refer to RFP Section 2.0.

Question 13: Is UConn expecting the bidders to include AI tools and technologies that their Researchers would like access to on this platform for: AI model training and tuning; building data lakes and data fabrics; and governing AI projects and models?

Answer 13: Refer to Answer 5 above.

Question 14: Does UConn expect to need the full capacity of 400 GPUs + 400 AIAs on day 1?

Answer 14: The proposer will articulate the proposed project plan and timeline in Question Group 2.2. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 15: The environment with 400 GPUs + 400 AIAs represents the upper end of compute‑intensive research and industry use cases—typically involving massive datasets, billion‑scale models, or real‑time multimodal processing. Share details on the use cases UConn intends to pursue.

Answer 15: Refer to RFP Section 2.0.

Question 16: Does UConn have data center space and power to run 400 GPUs + 400 AIAs?

Answer 16: Refer to Addendum 2, Question and Answer 20 and 48. The University does not have identified data center space.

Question 17: How many AI Researchers, participants & contributors does UConn expect to have day 1? Year 1, Year 2, Year 3?

Answer 17: Refer to Addendum 3, Question and Answer 5.

Question 18: UConn has stated that where the funds are being allocated from, is there an approved budget for this project?

Answer 18: Refer to Addendum 2, Question and Answer 29.

Question 19: Is the budget approval for the full 48 months?

Answer 19: Refer to Addendum 2, Question and Answer 29.

Question 20: What is estimated contract award date?

Answer 20: The anticipated schedule for the solicitation process is as stated in the RFP. The timeline for contract negotiation and commencement is to be determined. Refer to the RFP.

Question 21: What is the general timeline for approvals and signatures?

Answer 21: Refer to Answer 20 above. UConn contract negotiation and commencement timelines vary by complexity, subject matter, etc.

Question 22: What is UConn's target project start date in a best case scenario?

Answer 22: The proposer will articulate the proposed project plan and timeline in Question Group 2.2. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 23: Does UConn already operate HIPAA or NIST 800 171 compliant environments that require integration?

Answer 23: Refer to Addendum 2, Question and Answer 117.

Question 24: What identity provider does UConn use for RBAC/MFA?

Answer 24: The University uses Microsoft Entra for identity, and DUO for MFA.

Question 25: Will UConn provide early access to data center power/cooling/rack information?

Answer 25: Refer to Addendum 2, Question and Answer 20 and 48. The University does not have identified data center space.

Question 26: Are there any Managed Service delivery country restrictions or is it open for offshore service delivery?

Answer 26: Refer to Addendum 2, Question and Answer 70.

Question 27: Due to current market situation with component charges, price validity is at 7 days for hardware, whereas RFP asks for 180 days. How does UConn want the suppliers to address this disparity?

Answer 27: The University will not advise on the content of the proposal.

Question 28: What is the "Go Live" date for each enclave? a. General Research Enclave? b.

HIPAA Enclave? c. Federal Compliance Enclave?

Answer 28: The proposer will articulate the proposed project plan and timeline in Question Group 2.2. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 29: Can the vendor form a joint venture with other third parties post contract award?

Answer 29: Refer to the University’s template contract for this opportunity posted to the Buyer Attachments in HuskyBuy. Refer also to RFP Section 6.28 and 7.2. The University will not advise the proposer on the content of the proposal.

Question 30: Is there a preference for a public cloud service provider? (AWS, GCP, Azure)

Answer 30: Refer to Addendum 3, Question and Answer 20.

Question 31: Is there a requirement to keep the data in North American data centers or would data and technology be housed in UConn’s Information Technology Services (ITS) Data Center?

Answer 31: Refer to Answer 25 above. Refer also to Addendum 3, Question and Answer 21.

Question 32: Can the vendor propose solutions based on public cloud services without on prem resources?

Answer 32: Refer to RFP Section 3.0 – System Modality, “[t]he proposal may include three different modes: On premise, Cloud, and Hybrid (a combination of on premise and cloud) solutions.” The proposer will identify the proposed system modality or modalities in the response to Question 2.2.4.

Question 33: Can the on-premises resources be housed in third-party data centers or in close proximity to the campus?

Answer 33: Refer to Addendum 2, Question and Answer 50.

Question 34: What UConn internal resources will be available to coordinate set up connectivity? Will any data migration tasks be required?

Answer 34: Refer to RFP Section 3.0 – Technical & Programmatic Support, “[t]he proposal must also articulate proposer’s assumptions related to support by UConn personnel, if any.” Refer also to Addendum 2, Question and Answer 97.

Question 35: Is data currently stored in an object store like S3, NFS or other?

Answer 35: The RFP reflects the opportunity to provide a new service. Researchers may have files spread across many different storage areas currently.

Question 36: Will this network be bridged with the university network?

Answer 36: Refer to Addendum 3, Question and Answer 30.

Question 37: Which cloud resources are currently in use at UConn?

Answer 37: Refer to Addendum 3, Question and Answer 20.

Question 38: How do the resources access the cloud? Through a VPN - Vendor VPN or UConn

VPN?

Answer 38: Cloud connectivity varies depending on the service.

Question 39: Is any network connectivity required between the university data center and the vendor's data center?

Answer 39: Refer to Addendum 3, Question and Answer 30.

Question 40: Is integration with UConn Active Directory or a VPN necessary for authentication and authorization?

Answer 40: Yes, the service should integrate with the University’s IAM services.

Question 41: Which primary objective should vendors optimize for in Year 1? a) Maximum compute capacity, b) Predictable annual operating cost c) Phased growth aligned to funding availability?

Answer 41: Refer to Addendum 3, Question and Answer 3. Refer also to Addendum 2, Question and Answer 26.

Question 42: Is there an internally approved budget range for Year-1 operations that vendors should consider when sizing the solution?

Answer 42: Refer to Addendum 2, Question and Answer 29.

Question 43: Does UConn prefer steady-state annual pricing, or a front-loaded investment with reduced operating costs in Years 2–4?

Answer 43: In the competitive process, the University will not advise the proposer on the pricing proposal.

Question 44: Does UConn have a preferred deployment model (vendor-hosted, cloud-hosted, on-prem, or hybrid), or should vendors propose an optimized model?

Answer 44: Refer to Addendum 3, Question and Answer 7.

Question 45: If hybrid is preferred, which workloads or data types must remain on-prem (if any)?

Answer 45: Refer to Answer 44 above.

Question 46: Are there institutional or state policies restricting off-campus hosting for specific data classifications (e.g., HIPAA or federal data)?

Answer 46: Refer to Addendum 2, Question and Answer 56. Refer also to Addendum 3, Question and Answer 21.

Question 47: Does the ~400 GPU requirement represent Day-1 capacity, a Year-1 peak, or a multi-year growth target?

Answer 47: Refer to Answer 14 above.

Question 48: What average GPU utilization range should vendors assume for pricing purposes (e.g., 30%, 50%, 70%)?

Answer 48: Refer generally to RFP Section 2.0. Average GPU utilization range is undetermined.

Question 49: Does UConn anticipate burst-heavy or deadline-driven workloads (e.g., semester cycles or grant deadlines)?

Answer 49: Refer generally to RFP Section 2.0. Workloads are undetermined.

Question 50: Will UConn implement chargeback or showback mechanisms to influence or manage GPU consumption?

Answer 50: Refer to Addendum 2, Question and Answer 88.

Question 51: Does the ~15PB estimate represent total raw capacity, or a mix of active, cold/archive, and replicated backup data?

Answer 51: The storage capacity is useable, to include a mix of front-end cache and slower (object) storage.

Question 52: What approximate percentage of storage must be high-performance versus standard object or archival tiers?

Answer 52: It is left to the discretion of the proposer to propose a storage solution that it believes best fulfills UConn’s needs in the proposal.

Question 53: For regulated enclaves (HIPAA and Federal), does UConn expect dedicated physical infrastructure per enclave or logical segregation on shared infrastructure?

Answer 53: Refer to Addendum 2, Question and Answer 91.

Question 54: For the Federal enclave, is NIST SP 800-171 sufficient, or should vendors anticipate future FedRAMP alignment and associated artifacts (e.g., SSP, POA&M)?

Answer 54: Refer to Addendum 2, Question and Answer 75.

Question 55: Does the 4-hour response SLA apply to all environments, or only production and regulated workloads (with relaxed SLAs for non-regulated research)?

Answer 55: Refer to RFP Section 3.0 – Hardware and Software Needs, the identified minimum is “… 24/7/365 support with a 4-hour onsite response time for physical equipment issues, and a 4-hour response time for cloud service incidents.” The 4-hour response time is for all infrastructure.

Question 56: What categories of AI workloads does UConn expect this platform to primarily support? a) Inference-only (model serving, experimentation, applied research) b) Fine-tuning of existing foundation models c) Full model training from scratch d) A mix of the above?

Answer 56: Refer to Answer 49 above.

Question 57: If model training or fine-tuning is anticipated: What proportion of total GPU time should vendors assume will be allocated to training vs inference? Are training workloads expected to be continuous, episodic, or grant-driven? Does UConn expect vendors to provide training orchestration frameworks (e.g., distributed training, checkpointing, experiment tracking), or only the underlying infrastructure?

Answer 57: Refer to Answer 49 above.

Question 58: Where does UConn expect vendor responsibility to stop in the AI stack? Clarify expectations across the following layers: Physical infrastructure & facilities; GPU / accelerator hardware lifecycle; Cluster management (schedulers, orchestration, monitoring); Model hosting & inference endpoints; Model optimization, tuning, or performance engineering; Model governance, evaluation, and validation.

Answer 58: Refer to RFP Section 3.0 – Technical & Programmatic Support, “[t]he awarded proposer will run the AI resource as a managed service, ensuring stability, security, and performance on a 24/7 basis. Support for the entire stack is expected.

Question 59: Does UConn expect the vendor to support student-, faculty-, or researcher-developed models in production-like environments, or only shared/common models?

Answer 59: The University will not advise the proposer on the content of the proposal. It is left to the discretion of the proposer to identify support capabilities in the proposal.

Question 60: How should vendors differentiate between “research-grade” and “production-grade” workloads? Are there environments where best-effort support is acceptable? Are certain use cases expected to meet uptime, latency, or reliability guarantees?

Answer 60: Refer to Answer 59 above.

Question 61: Does UConn anticipate external-facing or revenue-generating AI applications (e.g., public APIs, industry partnerships, commercialization)? If yes, should those be architected separately from academic research workloads?

Answer 61: Refer to RFP Section 3.0 – Value-Added Considerations – Future Collaboration, “[p]roposers are invited to include in the proposal any value-added features or additional services or approaches to assisting the university in realizing the [listed] goals”, including applied AI commercialization. The University will not advise the proposer on the content of the proposals.

Question 62: Who owns prioritization and arbitration when GPU demand exceeds capacity?

Central IT; AI Institute leadership; Individual departments or PIs?

Answer 62: Prioritization and arbitration management has not been pre-determined.

Question 63: What identity and access management model does UConn expect? Integration with existing IAM (NetID, AD, federated identity); Role-based vs project-based access; Time-bound or grant-bound access policies.

Answer 63: Refer to Answer 40 above. The proposer will be expected to integrate with UConn’s IAM services (Microsoft Entra) and multifactor authentication (DUO).

Question 64: Should proposers assume responsibility for implementing usage quotas, fairness policies, or job preemption rules?

Answer 64: The University will not pre-evaluate the proposal. It is left to the discretion of the proposer to address service capabilities in the proposal.

Question 65: Without specific requirements and specific projects is a project plan and a key staffing plan pre-mature at this point?

Answer 65: The University will not pre-evaluate the proposal. The proposal will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 66: Some of the questions in the portal (i.e., 2.1.4) ask for attachments, but then state ("optional"). Please clarify if these types of questions require an attachment to submit the entire bid.

Answer 66: Response to optional question types are not required to complete the proposal submission process within the HuskyBuy portal.

Question 67: Does UConn plan to use any AI tools to evaluate proposal responses? If so, could you let proposers know which tool you plan to use?

Answer 67: No.

Question 68: What is the expected compute usage per day per user for each of the 3 required enclaves?

Answer 68: Refer generally to RFP Section 2.0. Compute usage per day per user is undetermined.

Question 69: What are the monthly storage requirements per month per user for each of the 3 required enclaves?

Answer 69: Refer generally to RFP Section 2.0. Monthly storage requirements per month per user is undetermined.

Question 70: If deployment in a hybrid environment would the services reside in the UConn cloud provider or would the vendors cloud account be utilized as a dedicated UConn environment?

Answer 70: Refer to Addendum 3, Question and Answer 20.

Question 71: Describe the maximum amount of pooled GPUs that any one AI job will need to be run across?

Answer 71: Unknown.

Question 72: Describe the type of specialized AI Accelerators and/or workload purpose.

Answer 72: Refer generally to RFP Sections 2.0 and 3.0.

Question 73: In Addendum #2, there is mention of 800 GPUs/AIAs across the three enclaves though the RFP only mentions 400. Can UCONN clarify the GPU/AIA breakdown across the three enclaves if they were to be physically isolated?

Answer 73: Refer to Addendum 2, Question and Answer 84. Refer also to RFP Section 3.0 – Hardware and Software Needs, “… 400 GPUs (latest generation) and a similar number of AIAs ….” (emphasis added).

Question 74: Will UCONN require air-gapped infrastructure to support current or net new federal grants?

Answer 74: Refer to RFP Section 3.0 – Required Features, compliance with all applicable security frameworks is required. The proposer will articulate how the design of the proposed solution supports the ability to meet applicable security frameworks in Question 2.2.12 in the HuskyBuy platform.

Question 75: Are there preferred GPU or accelerator families (for example NVIDIA H100, AMD MI300, or other AI accelerators), or do proposers have flexibility as long as performance objectives are met?

Answer 75: The University is conducting an open competition. Refer generally to Addendum 2, Answer 94.

Question 76: Is the target of approximately 400 GPUs, 400 AI accelerators, and 30+ PFLOPS expected to be delivered at Day-1, or can this capacity be phased over the life of the contract?

Similarly, is the ~15PB storage requirement a Day-1 expectation or a projected capacity target over time?

Answer 76: Refer to Answer 14 above.

Question 77: Does UConn have any preferred hyperscaler relationships (AWS, Azure, GCP) that may influence evaluation, or are proposers free to recommend the most suitable platform?

Answer 77: Refer to Addendum 3, Question and Answer 20.

Question 78: Does UConn have existing on-premises data center capacity intended for this deployment, or should proposers assume a greenfield or vendor-hosted infrastructure?

Answer 78: Refer to Answer 16 above.

Question 79: Does UConn have a preferred deployment approach (on-prem, cloud, or hybrid), or should proposers recommend the optimal model based on requirements? If an on-prem or hybrid model is proposed, what data center capacity is currently available in terms of power, cooling, rack density, and physical space to support high-density GPU infrastructure?

Answer 79: Refer to Addendum 3, Question and Answer 7.

Question 80: Is UConn expecting proposers to include pre-built AI/ML platforms (such as Kubeflow, MLflow, Ray, Slurm, or similar), or is there flexibility to propose an open and modular software stack?

Answer 80: Refer to Addendum 3, Question and Answer 19.

Question 81: For foundation models, what is the preferred approach: Vendor-provided models, Open-source models or Models developed, owned, or governed by the University?

Answer 81: Refer to Addendum 2, Question and Answer 100.

Question 82: Is there preferred operating systems, container platforms, or orchestration standards currently in use at UConn that should be aligned with? Also are there mandatory integrations with existing UConn systems (for example identity management, HPC schedulers, security monitoring, or logging platforms) that proposers should plan for?

Answer 82: Refer to Answer 40 above, the service should integrate with the University’s IAM services. Refer also to Addendum 2, Question and Answer 100.

Question 83: Is UConn seeking a fully vendor-managed AI platform with 24×7 operations, or a co-managed model with clearly defined responsibilities between the vendor and UConn IT/research teams?

Answer 83: Refer to RFP Section 3.0 – Technical & Programmatic Support, “[t]he awarded proposer will run the AI resource as a managed service, ensuring stability, security, and performance on a 24/7 basis. … The proposal must also articulate proposer’s assumptions related to support by UConn personnel, if any.”

Question 84: Are technology refresh cycles (such as GPU upgrades) expected to occur at fixed intervals, or should these be proposed as optional upgrades over the contract term?

Answer 84: In the competitive process, the University will not advise the proposer on the content of the proposal. It is left to the discretion of the proposer to describe the proposed approach to refresh cycles (hardware and software) in the proposal.

Question 85: Does UConn prefer pricing to be presented as a fixed annual managed service, usage-based consumption, or a hybrid commercial model? Also are there indicative budget ranges or funding constraints for the base AI environment versus the incremental HIPAA and Federal Compliance enclaves?

Answer 85: Refer to Answer 43 above.

Question 86: Should periodic technology refreshes be included within the base pricing, or treated as separately priced future enhancements?

Answer 86: Refer to Answer 84 above.

Question 87: Within the 20% weighting for “value-added considerations,” is UConn more focused on research collaboration, workforce development and training, or ecosystem and commercialization outcomes?

Answer 87: The University will not pre-evaluate the proposal. The proposal will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 88: For the General Research, HIPAA, and Federal Compliance enclaves, does UConn expect these environments to be physically isolated, logically isolated, or a combination of both?

Answer 88: Refer to RFP Section 3.0 – Required Features, “[e]ach Enclave must be logically segregated (and physically segregated where appropriate).”

Question 89: Does UConn prefer these enclaves to be fully vendor-managed, jointly managed with UConn IT, or managed by UConn with vendor support?

Answer 89: Refer to Answer 83 above.

Question 90: Should the HIPAA and Federal Compliance enclaves be delivered as fully compliance-ready environments (including controls, documentation, and operational processes), or is the vendor expected to provide compliant infrastructure while UConn retains governance responsibility?

Answer 90: Refer to RFP Section 3.0 – Technical & Programmatic Support, “[t]he proposal must also articulate proposer’s assumptions related to support by UConn personnel, if any.”

Question 91: For he Federal Compliance enclave, should proposers assume FISMA Moderate as the baseline, or is a different compliance level anticipated?

Answer 91: Refer to Addendum 3, Question and Answer 2.

Question 92: Is UConn currently HIPAA-certified, or is the expectation that the vendor enables and supports full HIPAA compliance readiness as part of the solution?

Answer 92: Refer to RFP Section 3.0 – Required Features, “[t]he Enclave must support HIPAA compliance, including technical, administrative, and physical safeguards.

Question 93: Should enclaves be designed as strictly isolated environments, or is controlled and governed cross-enclave data or workload movement expected?

Answer 93: Refer to RFP Section 3.0 – Required Features.

Question 94: Does UConn have a preference for physical versus logical segregation specifically for HIPAA and Federal Compliance workloads?

Answer 94: The solicitation utilizes a solutions-based model, rather than a prescriptive specifications-based approach. The proposer will articulate its proposed system architecture in the proposal. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 95: Is there a required deadline for the firm to achieve HIPAA compliance?

Answer 95: The proposer will articulate the proposed project plan and timeline in Question Group 2.2. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 96: Same question applies for NIST SP 800-171 and FISMA requirements?

Answer 96: Refer to Answer 95 above.

Question 97: Clarify the multi‑year vision for the AI Innovation Institute beyond the installation of the AI system (e.g., research enablement, commercialization, curriculum, grant partnerships)?

Answer 97: Refer to RFP Section 2.0.

Question 98: Does UConn have a prioritized list of AI use cases or research domains expected to be supported in years one through four?

Answer 98: Refer to RFP Section 2.0.

Question 99: How does UConn envision the role of the technology provider in shaping research programs, curriculum, and external partnerships?

Answer 99: The proposer, as the expert on its capabilities, will articulate the proposed value-added considerations in the proposal. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 100: Does UConn expect continuous use case development through the vendor team, or only infrastructure and platform operations?

Answer 100: Refer to RFP Section 3.0 – Value-Added Considerations – Future Collaboration.

Question 101: Should the proposal include AI governance and responsible AI frameworks, or does UConn plan to own this internally?

Answer 101: The University will not advise the proposer on the content of the proposal. Refer to Answer 5 above.

Question 102: Is there an expectation for dedicated onsite personnel, hybrid, or fully remote support? Are non-US resources permitted (e.g., FTE resources from a center in India)?

Answer 102: Refer to Addendum 2, Question and Answer 57.

Question 103: For cloud-based solutions, is the "4-hour onsite response time for physical equipment issues" requirement waived, or should proposers include physical on-campus presence regardless of modality?

Answer 103: On-campus presence is not mandated for a cloud solution.

Question 104: Are there specific grant deadlines, academic calendar dates, or research commitments requiring the system to be operational by a particular date?

Answer 104: No date has been identified by UConn.

Question 105: Does UConn have a preference among on-premises, cloud, or hybrid solutions?

If on-premises is preferred, please describe available datacenter facilities. Are there any policy restrictions that would preclude a cloud-only solution?

Answer 105: Refer to Addendum 3, Question and Answer 7.

Question 106: Does UConn already have existing Microsoft Fabric or Azure subscriptions, or should the vendor propose a new landing zone?

Answer 106: The University currently maintains a Microsoft Azure agreement, including a GCC High subscription. The service would be carved so the awarded proposer can support while maintaining centralized services such as authentication.

Question 107: What is the anticipated distribution of workloads across the three security enclaves (General Research, HIPAA, Federal Compliance) as a percentage of total compute?

How many concurrent users are expected per enclave?

Answer 107: Refer to Answer 49 above.

Question 108: Does UConn currently have an Azure Government tenant? What specific federal programs or sponsors (DoD, NIH, NSF, DOE) drive the NIST SP 800-171/FISMA requirement?

Answer 108: Refer to Answer 106 above.

Question 109: Does UConn Health currently have a HIPAA BAA with any cloud provider? Will the HIPAA Enclave handle actual PHI or only de-identified data?

Answer 109: UConn does not currently maintain a HIPAA BAA with a cloud provider. Refer to RFP Section 3.0 – Required Features, “ … HIPAA Enclave: An environment designed to support research involving Protected Health Information (PHI) and to enable the University to meet its obligations under HIPAA and associated regulations.”

Question 110: Beyond HIPAA and NIST SP 800-171/FISMA, are other compliance frameworks required (ITAR, EAR, FERPA, CMMC)?

Answer 110: Yes, UConn is subject to other compliance frameworks as a university.

Question 111: The RFP requires "immediate cloud services." How many GPUs must be immediately available? How much storage must be immediately available? How many users or research projects must be supported on Day 1? Which research domains need to be enabled immediately upon contract execution?

Answer 111: The proposer will articulate the proposed project plan and timeline in Question Group 2.2. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 112: Provide an overview of existing: Campus network topology; Identity systems;

Storage ecosystems; High‑performance computing resources.

Answer 112: The University’s identity system is Microsoft Entra.

Question 113: Are there existing solutions the Institute must integrate with (e.g., HPC clusters, OneDrive/SharePoint, GIS systems, clinical systems from UConn Health)?

Answer 113: It is left to the discretion of the proposer to articulate integration capabilities of the proposed solution in the proposal.

Question 114: What is the preferred approach for handling data ingress/egress between research groups?

Answer 114: The University has not pre-determined a preference.

Question 115: What identity provider(s) does UConn use (Shibboleth, Azure AD/Entra ID, Okta, on-premise AD)? Will external collaborators need access?

Answer 115: Refer to Answer 24 above.

Question 116: Does UConn currently have Azure ExpressRoute or equivalent private cloud connectivity? If not, what is the timeline for network infrastructure changes?

Answer 116: The University is currently moving forward with implementing Azure ExpressRoute.

Question 117: For the ~15PB storage requirement, what percentage is high-performance vs.

standard/archive?

Answer 117: Undetermined.

Question 118: Is there existing data to migrate into the new platform?

Answer 118: Refer to Answer 34 above.

Question 119: For the 30 PFLOPS (RPEAK) requirement, which precision metric should proposers use (FP64, FP32, FP16, or FP8)?

Answer 119: Refer to Addendum 2, Question and Answer 25.

Question 120: Is 30 PFLOPS (RPEAK) requirement combined across all enclaves?

Answer 120: Refer to Addendum 2, Question and Answer 36.

Question 121: Does UConn have AI ethics guidelines or responsible AI governance? Are certain AI use cases prohibited or require special approval?

Answer 121: Refer to RFP Section 2.0. Advancing the ethical use of AI will be a topic of the Institute. No prohibitions or approvals have been pre-identified.

Question 122: How does UConn define and measure success metrics over four years? Platform uptime? Number of use cases delivered? Research outputs?

Answer 122: Metrics have not been pre-identified.

Question 123: Should the proposal include: Quarterly or annual roadmap reviews?

Performance scorecards? User satisfaction surveys?

Answer 123: The University will not advise on the content of the proposal.

Question 124: Is an extension to the RFP submission deadline being considered?

Answer 124: An extension of the submission deadline is not anticipated.

Question 125: With a projected on-premises capability of approximately 30 petaflops, how does UConn envision its steady-state operating model once the environment is fully deployed and stabilized? Specifically: Do you anticipate reducing reliance on public cloud services over time? If so, what workloads would be prioritized for repatriation versus remaining in cloud environments? Are there cost, performance, or governance drivers shaping this long-term strategy?

Answer 125: The solicitation utilizes a solutions-based model, rather than a prescriptive specifications-based approach. The proposer will articulate its proposed system architecture, project plan, and timeline in the proposal to best fulfill UConn’s needs. Proposals will be evaluated in accordance with the Evaluation Criteria in the RFP.

Question 126: Does UConn have a preferred weighting or prioritization across deployment models (on-premises, public cloud, or hybrid) as part of this initiative? Additionally: Is there a defined future-state architecture the university is targeting? What does success look like from an infrastructure and operational standpoint over the next 2–5 years?

Answer 126: Refer to Addendum 3, Question and Answer 7.

Question 127: From an implementation perspective, does UConn expect the balance of resources to evolve between initial Day-1 operations and later steady-state phases (for example, Day-90 or Year-1)? For example: Should partners assume heavier cloud utilization during early phases, followed by increased on-prem adoption as capacity ramps? Are there anticipated shifts in workload placement as the environment matures?

Answer 127: Refer to Addendum 3, Question and Answer 7.

Question 128: Beyond the deployment model (on-premises versus cloud), does UConn have defined expectations for the core operational personas required to support this AI environment? Specifically: What roles or functions does the university anticipate needing (for example, platform administrators, HPC engineers, data scientists, MLOps engineers, security operations, research support staff)? What level of proficiency is expected for each persona at go-live versus steady state? How does UConn envision day-to-day operational ownership and responsibility being distributed across these roles? Are internal teams expected to retain primary ownership, with partners augmenting specific functions?

Answer 128: Refer to Addendum 2, Question and Answer 69. Refer also to RFP Section 3.0 – Technical & Programmatic Support.

End of Question and Answer Addendum 5