Protest alleging that agency applied unstated technical requirements is denied. The agency found the protester’s proposed coding solution unacceptable because the solution failed to launch in Amazon’s cloud computing platform. The protester argued that its solution failed because of an undisclosed limitation in Amazon’s platform. The protester further argued that when the agency learned of this limitation, it should have revised the solicitation to allow new coding solutions. But GAO found that the protester’s solution failed because of the protester’s own coding choices. The protester chose to use a virtual server that had limited functionality on Amazon’s cloud. The protester’s attempts to blame the consequences of its own choices on Amazon or on unstated evaluation criteria were unavailing.
The Department of Homeland Security published a solicitation for integrated data analytics, artificial intelligence/machine learning, and development, security, and operations services. The solicitation prescribed a two-step evaluation process. In the first step, the solicitation presented a problem statement. Offerors were required to respond to the problem with a coding submission. The coding submission was supposed to launch an interactive notebook that would run within Amazon’s Web Services. The three offerors with the highest rated coding submissions would then move on to the second step of the evaluation in which DHS would evaluate how well the offerors performed in technical demonstrations.
Accenture submitted a proposal in response to the solicitation. As part of its coding submission, Accenture chose to use Amazon’s SageMaker instance. An instance is a virtual server for running applications on the Amazon infrastructure. Amazon’s cloud computing systems is divided into various regions depending on geographic location. Accenture’s solution specified that its solution should run on Amazon’s us-east-1 region. But when DHS attempted launch Accenture’s coding solution, it received an error message stating that the SageMaker instance was not available in us-east-1 region. Because DHA was unable to launch Accenture’s solution, it found Accenture’s proposal technically unacceptable.
Accenture filed a protest contending that Amazon’s limitations on the availability of the SageMaker instance in the us-east-1 region was an unstated technical requirement. Accenture argued that because there was no public documentation of the SageMaker limitation in the us-east-1 region, no offeror could have known not to propose the SageMaker in that region. Thus, the argument continued, when DHS became aware of the limitations on the SageMaker instance, it should have revised the solicitation to allow offerors new coding submissions.
GAO did not find this compelling. The solicitation did not require offerors to use the SageMaker instance. What’s more, DHS had no obligation to ensure that an offeror’s chosen instance was available in the in region they selected for their coding submission. The failure of Accenture’s solution was not the due to unstated criteria; rather, it was the result of Accenture’s own coding choices. Accenture’s attempt to blame its unacceptable rating on Amazon’s undisclosed limitations on the SageMaker instance instead of its own coding choices was nothing more than disagreement with the agency’s judgment.
Accenture is represented by Amy L. O’Sullivan, Olivia Lynch, and Stephanie L. Crawford of Crowell & Moring LLP. The agency is represented by John E.Cornell of the Department of Homeland Security. GAO attorneys Christopher Alwood and Christina Sklarew participated in the preparation of the decision.
