Solution Submissions are due by November 13, 2026!
NAVAIR/NAWCAD
The Central Florida Tech Grove, in partnership with NAVAIR/NAWCAD, is launching a prize challenge to identify innovative solutions for advanced acquisition documentation analysis. The challenge seeks intelligent recommender systems that leverage advanced data science, natural language processing, and machine learning to analyze procurement documentation, identify common acquisition requirements, and recommend strategic contract vehicles that improve efficiency and streamline the acquisition process.



Overview
NAVAIR executes tens of thousands of procurement actions annually. While currently leveraging sophisticated tokenization and embedding models for Retrieval-Augmented Generation (RAG), there is a critical need to evolve beyond simple document retrieval to unlock the analytical power of data cluster analysis by grouping procurement documents based on their content. The purpose of this competition is to build a recommender system that can analyze sources sought notices documents and open-source information (such as Congressional testimony) and return a candidate list of contract vehicles for consolidation.
Problem Statement
The challenge aims to develop a fully documented, validated algorithmic methodology capable of evaluating, identifying, and recommending common requirements from sources sought notices documents as well as open-source information such as Congressional Testimony (GFI list will be provided to approved participants). The algorithm will automatically produce candidate identification and recommendations for strategic vehicles. This includes consolidating similar requirements into larger vehicles or identifying new strategic contracting vehicles based on future capabilities.
Strategic Contract Vehicles are enterprise scale contracts that allows a large amount of requirement owners to procure items/commodities under the same base contract using streamlined procedures. While strategic vehicles can encompass vast requirements, they must maintain the ability to levy consistent and performance scope requirements to vendors. The benefits of this approach include reduced duplication of effort, improved buying power, maximize contract ceilings, and improve speed of acquisition.
Today’s DoD systems contain sparse databases of information with no systematic way to determine common requirements at scale. For example, if VX-XX has a requirement for engineering services on the flight line, the requirement owner has to establish their own contract. This is one of many locations that needs to buy the same service. A better approach would be to identify similar contracting vehicles that can be utilized to fulfill this requirement. In the current ecosystem, the process relies on an individual with knowledge of existing systems, requirements, and contracting vehicles. In order to streamline acquisition processes, there is a need to automatically identify contracts suitable to be used as strategic vehicles.
Benefits:
The primary benefits include increased buying power, reduction of duplication significant process efficiencies, speed of acquisition, and cost savings that directly support DoD acquisition reform efforts.
Critical Technical Criteria:
Deployment Environment: The solution must be designed for deployment within a U.S. Government-owned and operated Cloud environment, accredited for Impact Level 4 (IL4) data.
System Architecture: The core analysis and identification methodology must ensure repeatable results (produces same top 5 results 95% of the time). The final summarization component must consist of visualizations to articulate the solution. Database schema and ETL process must be provided if database is used.
Visualization: Must include a visual representation of the analysis method utilized and results.
Summary Metrics: Candidates must provide a summary metric for how they evaluated algorithm performance.
Dependencies: Deliverables must be packaged in a Docker container with all external dependencies and cannot access external services outside of USN-approved models or microservices within an IL4 environment (i.e., no external industry-hosted custom models).
Prizes/Awards
Cash Prizes: $150K Prize Challenge pool.
1st Place Prize: $100,000
2nd Place prize: $50,000
Exposure to leaders across Team Opportunities Orlando and other branches of the U.S. Department of Defense (DoD) responsible for acquisition of development and deployment of training solutions.
Opportunities to learn from and collaborate with other industry, government, and academic leaders and partners in this space
Potential for follow-on opportunities such as Other Transaction Agreements (OTAs), Cooperative Research and Development Agreements (CRADAs), and others.
Timeline
Prizes
in total cash prizes
$100,000 awarded to the top finalist; $50,000 awarded to the second place finalist
Additional Considerations-
How to Enter
Interested participants must complete the official Pre-Screening Questionnaire by the published deadline.
The questionnaire will be used to evaluate and verify:
Submit Pre-Screening Questionnaire >
Participants selected to advance will receive access to the Government Furnished Information (GFI) and the official solution submission portal. Completion of the questionnaire does not guarantee advancement.
Approved participants will receive access to the Government Furnished Information (GFI) and the official solution submission portal.
The Phase 2 submission must include an initial technical package containing:
Code must be fully reproducible, utilize Docker, and be capable of operating in an IL4 environment.
The submission must not consist solely of a link to a website.
Government technical evaluators will assess the submitted solutions using the published judging criteria. Semifinalists will be selected to advance to Phase 3.
Government Furnished Information (GFI)
GFI will be provided only to participants approved to advance beyond the Pre-Screening Questionnaire.
Up to seven semifinalists will be invited to participate in the final Demo Day.
Semifinalists must provide final presentation materials in accordance with the instructions and deadline provided by Tech Grove.
Final presentation materials must include:
During Demo Day, each semifinalist will receive a validation dataset similar to the test dataset. The solution will be evaluated based on its ability to identify the manually identified candidates contained within the validation set
Judging
Phase 1: Pre-Screening Questionnaire
Complete the Pre-Screening Questionnaire to be considered for Phase 2 of the Prize Challenge. Selected participants will receive access to the Government Furnished Information (GFI) and instructions for submitting their solution.
Phase 2: Government Furnished Information (GFI) & Solution Submission
Participants selected to advance from Phase 1 will receive access to Government Furnished Information (GFI) along with instructions for submitting their solution. Your solution must be submitted in the format outlined below and will be evaluated using the published judging criteria. Only complete submissions that satisfy all submission requirements will be evaluated. Semifinalists will be selected to advance to the Final Demo Day.
Government Furnished Information
GFI will be provided only to participants approved to advance beyond the Pre-Screening Questionnaire.
Participants must submit the following:
Initial Technical package (algorithm documentation, codebase, packaging requirements) followed by Final Presentation Materials for Demo Day. Code should be provided and fully reproducible. It should utilize Docker, and be capable of running in an IL4 environment.
It should NOT be a link to a website.
The final presentation (Phase 3) should include:
Solution Validation: 50 Points
Validation will be performed in the initial evaluation by the Government technical team, and the remaining verification will be performed during the live demo day.
Initial Technical Evaluation (30 Points):
Each data point in the dataset with a correct prediction will be granted one point. Correctness is defined as a result being within a group of 20 results predetermined as correct by the Procurement Group Innovation Lab (PGIL) on the validation set. Solutions that provide excessive cluster groupings will be negatively impacted in the Alignment criteria.
5 points granted if container can be replicated. This means the solution can be distributed across multiple containers without affecting the results. Replication must demonstrably improve performance.
0 points granted if container cannot be replicated without affecting results.
3 points granted if the solution runs on 1 core, 2GB RAM
2 points granted if the solution runs on 4 cores, 8GB RAM
1 point granted if the solution runs on 8 cores, 16GB RAM
0 points granted if the solution requires more than 8 cores, 16GB RAM
2 points granted for a solution that does not use an LLM
1 point granted if a solution uses an LLM and spends less than 50k tokens per retrieval
0 points granted if a solution uses an LLM and spends more than 50k tokens per retrieval
Live Demo Day (20 Points)
During the live demo day, the researcher will be given a verification set similar to the GFI test set to demonstrate their solution. Solutions will be evaluated in the live demo day by their ability to identify five manually identified candidates from the verification test set.
4 points granted for each correctly identified candidate
The additional datasets must be processed and candidates for recommendation must be provided in 30 minutes. This is a time requirement, not merely a validation requirement. Solutions that do not meet the 30-minute time requirement will not be considered.
Solution Alignment: 50 Points (20 points during the Initial Technical Evaluation and 30 points during the Live Demo Day)
The evaluators will specifically review the submission for its effectiveness to provide meaningful clusters representing alignment to the definition of Strategic Vehicles in the specific domain of Naval Aviation. A Strategic Vehicle is defined as an enterprise scale contract that allows a combined group of requirement owners to procure items/commodities under the same contract using streamlined procedures. While strategic vehicles can encompass vast requirements, they must maintain the ability to levy consistent performance scope requirements to vendors and exist within FAR/DFARS regulations governing USG contracts. The targeted benefits of this approach include reduced duplication of effort, improved buying power, maximize contract ceilings, and improve speed of acquisition.
Phase 3: Demo Day and Final Judging
Following the Government's Phase 2 evaluation, up to seven semifinalists will be selected and invited to participate in the Final Demo Day.
Semifinalists will receive instructions for submitting their final presentation materials prior to the event.
Live Demo Day (20 Points)
During the live demo day, the researcher will be given a verification set similar to the GFI test set to demonstrate their solution. Solutions will be evaluated in the live demo day by their ability to identify five manually identified candidates from the verification test set.
4 points granted for each correctly identified candidate
The additional datasets must be processed and candidates for recommendation must be provided in 30 minutes. This is a time requirement, not merely a validation requirement. Solutions that do not meet the 30-minute time requirement will not be considered.
Solution Alignment: 50 Points (20 points during the Initial Technical Evaluation and 30 points during the Live Demo Day)
The evaluators will specifically review the submission for its effectiveness to provide meaningful clusters representing alignment to the definition of Strategic Vehicles in the specific domain of Naval Aviation. A Strategic Vehicle is defined as an enterprise scale contract that allows a combined group of requirement owners to procure items/commodities under the same contract using streamlined procedures. While strategic vehicles can encompass vast requirements, they must maintain the ability to levy consistent performance scope requirements to vendors and exist within FAR/DFARS regulations governing USG contracts. The targeted benefits of this approach include reduced duplication of effort, improved buying power, maximize contract ceilings, and improve speed of acquisition.
Rules
Eligibility Requirements
Participants may be individuals or teams from industry, academia, state/local government, or non-profit entities. Eligibility is subject to verification during each Phase and before award of any prize. To be eligible to win, all individual participants or members of a participating team must meet the following eligibility requirements:
1. Participants must have complied with all requirements set forth in the prize challenge.
2. All individuals must be at least 18 years of age and a citizen of the United States.
3. Except for employees appointed under the STEM Student Employment Program or other student employment programs, Federal employees and support service contractors are not eligible to participate.
4. Participants may not use Federal funds to support participation.
Rules:
NAVAIR/NAWCAD reserves the right to cancel, suspend, and/or modify the Challenge, or any part of it, for any reason, at their sole discretion.
Terms and Conditions
Representations:
Upon submission, the Participant hereby represents and warrants that:
1. It is the sole author and copyright owner of the submission; that the submission is an original work of the Participant and that the Participant has sufficient rights to use and to authorize others, including Tech Grove and,
Conditions; and
Data Rights and Marking
Relationship of the Parties
Nothing contained in these Terms and Conditions is intended to create or constitute a relationship between Tech Grove,
Participant Liability and Insurance
Disputes
Follow-On Activities
This Open Call Announcement is considered to have potential for further efforts that may be accomplished via FAR-based contracting instruments, Other Transaction Authority (OTA) for Prototype Projects 10 USC 4022 and Research 10 USC 4021, Prizes for advanced technology achievements 10 USC 4025, and/or Prize Competitions 15 USC 3719. If a prototype OTA is awarded and considered successfully completed, follow-on production may be pursued in accordance with 10 USC 4022(f). The public open call announcement made at https://www.challenge.gov/ is considered to satisfy the reasonable effort to obtain competition in accordance with 10 USC 4025(b), 15 USC 3719(e) and 10 USC 4022(b)(2). Any FAR-based actions will follow announcement procedures per FAR 5.201(b) accordingly.
Intellectual Property Considerations
By participating in this competition and submitting materials in any round, participants grant the U.S. Government and its Partners access to submitted prototypes, documents, and materials for evaluation purposes during the Prize Challenge. Participants retain intellectual ownership and publishing rights to all submitted materials.
Prize Challenge winners grant the U.S. Government and its Partners permanent access to the provided prototypes, documents, and materials, with the provision that the US Government and its Partners may store and copy the documents in perpetuity. Non-winning submissions shall not be retained or used by the Government beyond Prize Challenge evaluation and archival requirements.
Excluding any material considered to be co-produced with the US Government, the US Government and its Partners agree not to share submissions with other entities, including files and tradecraft concepts.
About the Central Florida Tech Grove
The Central Florida Tech Grove is a Department of Defense innovation hub established in 2020 under a Partnership Intermediary Agreement between the Naval Air Warfare Center Training Systems Division and the University of Central Florida Research Foundation. The Tech Grove was founded as a multi-service innovation hub with funding from the Navy, Army, Air Force, and Marine Corps training commands as well as the Office of Naval Research and the Office of the Secretary of Defense. With a three-pronged mission to (1) grow the defense industry base, (2) facilitate technology transfer and transition, and (3) solve problems, the Tech Grove executes a wide range of initiatives that bring together entrepreneurs, industry, academia, and government from across the United States and internationally to develop innovative solutions for military human performance. www.centralfloridatechgrove.org
Satisfying Competition Requirements: Other Transaction Authority
This public announcement is an open call to small businesses and non-traditional defense contractors seeking innovative, commercial technologies proposed to create new DoW solutions or potential new capabilities fulfilling requirements, closing capability gaps, or providing potential technological advancements, technologies fueled by commercial or strategic investment, but also concept demonstrations, pilots, and agile development activities improving commercial technologies, existing Government-owned capabilities, or concepts for broad Defense application(s). As such, the Government reserves the right to award a contract or an Other Transaction for any purpose, to include a prototype or research, under this public announcement. The Federal Government is not responsible for any monies expended by the applicant before award and is under no obligation to pursue such Other Transactions.
This Open Call Announcement is considered to have potential for further efforts that may be accomplished via FAR-based contracting instruments, Other Transaction Authority (OTA) for Prototype Projects 10 USC 4022 and Research 10 USC 4021, Prizes for advanced technology achievements 10 USC 4025, and/or Prize Competitions 15 USC 3719. If a prototype OTA is awarded and considered successfully completed, follow-on production may be pursued in accordance with 10 USC 4022(f). The public open call announcement made at Challenge.gov is considered to satisfy the reasonable effort to obtain competition in accordance with 10 USC 4025(b), 15 USC 3719(e) and 10 USC 4022(b)(2). Any FAR-based actions will follow announcement procedures per FAR 5.201(b) accordingly actions will follow announcement procedures per FAR 5.201(b) accordingly.
Additional Links, Resources, and FAQs
Virtual Information Session Recording
The solicitation said that the Phase 2 submission must include an initial technical package containing the complete codebase. Our platform could solve the problem statement, but we aren't willing to expose the entire codebase because that is core IP. Is that a non-negotiable?
The guide is clear. You must expose your entire codebase as part of the competition. It is non-negotiable.
Could you clarify how you are defining "support service contractors" for eligibility item 3 please? I have a client who may be interested, but supports other parts of the Government (not DoW).
All federal government support service contractors are excluded from the Advanced Acquisition Documentation Analysis Prize Challenge.
Does the Government want a new system or enhancements to a current system? If latter, what is the current system called and what are its limitations?
The government is looking for a new system.
Is there a price range for the required solution?
See Prize Challenge page on Tech Grove website.
How many licenses are required?
Not applicable. This is a prize challenge for a Dockerised prototype, not a software purchase; no license count exists. The Government receives permanent access to the submitted image, source code and documents; the participant keeps ownership.
Are there any specific pain points that are drivers for this announcement?
See Prize Challenge page. The driver is duplicated contracting. The same service is bought separately at many locations because each requirement owner sets up their own contract; DoW databases are sparse and there is no systematic way to find common requirements at scale, so the work depends on individuals who happen to know the existing vehicles. Within Procurement Group Innovation Lab, the benefits are: fewer duplicate contracts, more buying power, faster acquisition, cost savings.
What constitutes one scored data point and its prediction: a document, requirement, contract, existing vehicle, cluster, or proposed consolidation? Are candidates ranked per input requirement or once for the entire corpus? Please provide the required identifiers and output schema. Is there a correct answer at all? or is it based on our metrics, if so can you provide some example questions and answers.
A scored data point in this case is a group or cluster. See schema provided.
This is an unsupervised learning problem. There is no single correct answer.
No example question/answers will be provided.
Does the group of 20 correct results represent acceptable answers per query or 20 targets across the dataset? How are top-five results, ordering, duplicate contract families, partial credit, unjudged candidates, abstentions, and fewer than five eligible results scored?
For the initial submittal, we are looking for 20 groups or clusters.
NAVAIR SMEs will determine how many points to award based on quality of responses.
Scoring algorithm is not publicly available.
What is the unit of a "correct prediction",a document, a requirements cluster, or a contract vehicle? Will participants receive labeled examples or other ground-truth data to validate their solution before submission?
See responses above.
There are no labeled examples or ground truth.
What exact output schema is required: a ranked top-five candidate list for each input record, cluster assignments, proposed strategic-vehicle groupings, or some combination?
See output schema provided.
What training examples, relevance judgments, explanatory rationales, and negative examples will be provided? Please identify when GFI becomes available and whether a separate validation set and public-source corpus may be used for development.
No examples will be provided.
GFI provided to approved applicants. This is the data competitors will use for initial submittal.
The GFI includes repeated solicitation numbers and multiple notice types. Are repeated numbers separate versions or duplicate records, and should non-sources-sought records be included in the evaluation?
Repeated solicitation numbers are reposts or updates of the same contracting opportunity, not separate opportunities; treat records sharing a solicitation number as one. All notice types in the GFI are in scope for evaluation.
What file types, document formats, and approximate data volume are included in the Government Furnished Information (GFI)?
See GFI Provided.
Will Phase 2 participants receive labeled examples, ground-truth mappings, or other reference data that may be used to develop and validate the algorithm before submission?
No.
Workload envelope. Please provide corpus and incremental-batch document counts, total bytes, page counts, languages, file types, and scanned-page percentage. Does the 30-minute clock include container startup, OCR, parsing, indexing, model fitting, clustering, visualization, and export?
See GFI Provided. Part of your task is to understand the corpus provided.
30 Minutes includes the entire run, from container start to output, including demonstration. It does NOT include Q&A.
Resource accounting. Are CPU and RAM tiers measured across all participating containers or per container? Are the database, coordinator, cache, sidecars, and approved model services included? Please specify CPU architecture, disk limits and throughput, swap policy, and cold-cache conditions.
CPU & RAM is measured on the total of everything the participant submits.
No approved model services will be included.
Architecture: x86-64 Disk: 20 GB
no throughput promise
Swap: none
Cache: cold start.
No Open Source LLM.
Must replication improve a single batch's completion time or only throughput across simultaneous jobs?
No this is not part of the scoring.
What baseline and resource allocation apply? Does "same top five" mean identical membership or order, and must this hold across hardware types, restarts, input order, and worker counts?
Replication = two copies run side by side without breaking (pass/fail)
Repeatability = three runs in a row give the same membership, not order)
No throughput or speed-up claim is asked for or scored.
Does the 30-minute clock include data retrieval, document parsing, and index creation, or may participants submit a prebuilt index? Will all teams be tested on the same hardware configuration?
See above question regarding 30 minute time.
How will the 95% top-five repeatability requirement be measured? For the replication points, must multiple containers produce identical recommendations while demonstrating a measurable performance gain?
We will run your container 3 times. Group Membership must remain the same at least 95% of the time.
For replication, see above.
Performance gain is not part of the scoring.
For the replicability criterion, how will multiple containers be orchestrated and measured, and must the submission expose a specific interface or support partitioned input?
Two instances of the container are run on the same input; each must complete and write its own output. No specific interface, orchestration support, or partitioned input is required.
For the requirement that the core methodology produce the same top-five results 95% of the time, how will repeatability be tested and over how many runs?
See above.
For the 30-minute processing requirement, is there an expected or standardized hardware environment and approximate validation-dataset size participants should design toward?
See scoring on Prize Challenge page.
Validation dataset to be provided.
Models and preprocessing. Are supplied embeddings available with model/version metadata? Do local encoder-only embedding models count as LLM use? Are preprocessing and model fitting before the timed evaluation allowed, and how are tokens counted for any permitted model service?
No embeddings are furnished; you bring your own model inside the image.
A model that outputs numbers (sentence or document embeddings) is not an LLM; only a component that generates text counts. Pre-fitting is allowed, but everything must be inside the image and the full run must finish in 30 minutes from the run command. Tokens = prompt plus completion, summed over the full run.
Please identify the Government-operated hosting environment, permitted base images and packages, container runtime, storage interfaces, identity integration, logging expectations, approved microservices, and applicable authorization and data-handling requirements.
Please adhere to guidance provided in competition website. Evaluation does not happen in IL4. The judging machine is air-gapped, no network, no reachable services, no credentials; the container reads from a mounted input folder and writes to an output folder. Nothing in the eventual IL4 environment is available during evaluation, and no approved microservices exist to call. Competitors need not own or accredit an IL4 environment.
Should the solution call Congress.gov live during evaluation, or may it use a locally cached dataset? Given the IL4 connectivity limits, will the Government provide an approved data snapshot, API proxy, or other access method?
The container may not call Congress.gov during evaluation. Any Congress.gov data a solution uses must be included in the container.
Can the application itself be running in a native Linux web hosting environment with supporting systems like Nimio, Keycloak, PostgreSQL, Gotenburg, using separate Docker containers?
No. Evaluation runs Docker containers with no network access. Anything your solution depends on must be packaged inside your container/s.
What exact container input contract will evaluators use (invocation command, mounted paths, file types, and required schema)?
The container will start with a single run command, reading the GFI spreadsheet from a mounted input folder and writing its output files to a mounted output folder. The GFI is the only input the Government supplies; any Congress.gov data a solution uses must already be inside the container. The container should accept the input and output locations as arguments or environment variables rather than assume fixed paths.
During evaluation, will Congress.gov data be supplied as an offline snapshot/input, or is network/API access expected or permitted from the container?
See Previous Questions.
Does the submitted Docker container need to operate completely offline, with all models and dependencies packaged locally, or may it call specific Government-approved models/microservices during evaluation?
The container must operate completely offline with all models and dependencies packaged inside it. No external service is reachable during evaluation. We recommend to error on the side of caution if using models such as Amazon bedrock as the test environment can change.
Are pretrained open-source embedding/NLP models permitted if their weights are packaged inside the container and no external service is contacted?
Yes.
Contract relationships. Will metadata distinguish base vehicles, task orders, modifications, solicitation versions, and associated SOW/PWS/CDRL attachments? Are stable contract, vehicle, owner, organization, and document identifiers included?
Do not focus on the metadata.
Vehicle availability. Will authoritative scope, ordering eligibility, ordering period, ceiling and available capacity, geographic limitations, security requirements, set-aside status, and contract-office contacts be supplied? How should unknown or stale metadata affect recommendations?
GFI is all you will be provided with.
Consolidation criteria. What criteria make similar requirements operationally suitable for consolidation? Please identify acceptable scope variation, service-level differences, timing constraints, transition considerations, and applicable small-business and competition review factors.
Part of your task is to determine what criteria make similar requirements operationally suitable for consolidation. You will not be provided with further guidance on this matter. No operational consolidation criteria are published in either guide. In your submission, you can explain your own reasoning on these points.
Future capabilities. How are recommendations derived from testimony and news evaluated? What planning horizon, source-date cutoff, provenance standard, and corroboration level apply? Must a proposal map to funded requirements, or may it represent an emerging opportunity?
There are no published expectations on planning horizon, source-date cutoff, provenance standard or corroboration level is defined; or emerging opportunity proposals. In your submission, you can explain your own reasoning on these points.
Is there a list of the USN approved/supported models or microservices that we can use?
Below are the Amazon Bedrock models we have available. This list is subject to change.
Amazon:
Amazon Nova Multimodal Embeddings
Titan Text Embeddings V2
Meta:
Llama 3 70B Instruct
Llama 3 8B Instruct
Nvidia
NVIDIA Nemotron Nano 12B v2 VL BF16
NVIDIA Nemotron 3 Super 120B A12B
NVIDIA Nemotron Nano 9B v2
Nemotron Nano 3 30B
OpenAI
gpt-oss-120b
gpt-oss-20b
xAI
Grok 4.6
Live validation. How do the five manually identified candidates map to the 10 live-demo points? Are these five documents, vehicles, or clusters? Please distinguish the processing time allowance from the presentation, demonstration, and Q&A time allowance.
You are to provide 5 groups or clusters. Each will be worth 4 points.
See time allowance above.
For approved Phase 2 participants who cannot attend the live information session, will a recording, slides, written Q&A, or other session materials be distributed afterward?
Recording of Info session and Q&A to be provided on Tech Grove Prize Challenge website.
Reach out to our team by completing the form below.
© Copyright Central Florida Tech Grove. All Rights Reserved