SBIR/STTR Research Award
The Attic's GenLLM - Creation of Datasets and Fine-tuned Models in a No-code Platform
Award amount
$130,654Program and phase
SBIR · Phase IAward year
2025Award details
- Awarding agency
- Department of Defense
- Branch
- Office of the Secretary of Defense
- Recipient company
- THE ATTIC AI, INC.
- Proposal award date
- 2025-08-29
- Contract end date
- 2026-02-28
- UEI
- WBABLTXTSLJ6
- Solicitation number
- 25.4
- Topic code
- OSD254-P001
Research abstract
The Attic AI presents GenLLM, a no-code platform that enables the DoD to create custom AI models from their own data within days or hours. GenLLM is a commercial system that we propose to apply to the DOD and extend beyond its current capabilities to include agent creation and workflow, enabling anyone to build sophisticated, automated AI models and processes through a simple drag-and-drop interface.áGenLLM turns your data into datasets, creates a constitution, and then trains a model for you in a no-code platform.Data Transformation and Synthetic Data Generationôit ingests data from various formats, turning it into structured datasets ready for training. When necessary, the system can generate synthetic data to enhance robustness and diversity in training.öAI Constitution for Compliance and CustomizationThe AI constitution can embed ethical and operational guidelines directly into the models, ensuring that every action is aligned with mission standards and security protocols.öCustom Model Training and DeploymentôWith the datasets created, users can train a custom AI model without coding. This will enable specific models to be created for any DOD need and mission.Agentic Workflows for Task AutomationôIn addition we will integrate drag and drop agent creation enabling the automation of complex tasks like data validation, analysis, integration, and reporting. The drag-and-drop interface for creating AI agents makes it easy to implement into workflows.It offers unmatched deployment flexibility. Unlike other AI solutions limited to cloud environments, GenLLM enables models to be downloaded and deployed wherever theyĺre needed, including in austere or remote environments. And even using GGUFGenLLM is ready for DoD adoption. Itĺs currently at a TRL of 6, proven through user testing and we have deployed agentic workflows at a TRL of 8 in real-world applications. This Phase 1 seeks to apply these proven components to the DoD, creating a one stop AI solution for model creation, ensuring it fully aligns with mission requirements and lays the groundwork for broader deployment across defense operations.
Technology keywords
Not reported
Awards with the same topic code
Verified by an exact topic code match. This is not a text-similarity match.
- MyRadar Total AI Weather Forecasting Solutions
- AI for Airworthiness Certification
- FlightEdge: Real-time, Distributed AI/ML Model Training for Predictive Maintenance
- Dynamic, Adaptable Routing with Generative AI and LiDAR Data
- VOCAL: Voice Originality Cloning Across Languages