D. Mesa, J. Pisano, N. Pisano, M. Pitstick
SNA Software LLC, Florida, United States
Keywords: artificial intelligence, data management, data quality, PPM, project systems
The core issue in DoD digital systems that will facilitate utilization of more powerful technologies such as AI. We have developed an AI with the moniker "Occam" to use Generative AI methods, particularly Retrieval-Augmented Generation (RAG) and fine-tuning, to validate and interrogate project-focused schema files. The Occam AI platform, which is fully closed-source, autonomously validates the quality of these files, thereby making the judgment calls contained in those files readable and useable. Security and confidentiality are always at a premium for a platform that works with such sensitive defense applications, and a closed-source solution guarantees those high standards. Our unique approach tackles core problems in project management and data validation under the VAULTIS framework. It combines the output of several large language models (LLMs)—Ollama, ChatGPT, Claude, Google Gemini, Mistral, and Groq—which are integrated with Azure AI Studio and Python. This combination of models and technologies makes Occam AI a powerhouse for managing and verifying Integrated Program Management Analysis Report (IPMDAR) files and project management and systems engineering data; it does the job with impressive accuracy. The way it achieves this is by using SLIM (Structured Language Instruction Model) guidance.