ASAP (Advanced Semantic AI Platform)
AI-Powered Workspace for Streamlining Software Coding and Documentation
AI-Powered Workspace for Streamlining Software Coding and Documentation
Winner!
By Constantine Vassilev. July 19, 2024
I'm excited to announce another achievement in our journey! Our Advanced Semantic AI Platform (ASAP) has been named one of the 10 winners at the Microsoft Developers AI Learning Hackathon!
I was motivated to join the hackathon to explore the data infrastructure behind applications like OpenAI's ChatGPT, particularly the fusion of AI's speed with the dependability of traditional databases such as Azure Cosmos DB. With 30 years as a software engineer, I understand the challenge of searching through extensive databases for specific data, which can be a lengthy process. Using natural language for queries simplifies this search, cutting down on irrelevant data and increasing productivity.
This inspiration led me to develop the Advanced Semantic AI Platform (ASAP), a collection of AI-powered copilots designed to automate tedious coding and documentation tasks. The ability to perform precise searches with natural language and store data in a vectorized format in Cosmos DB was the perfect fit. During the hackathon, I discovered additional benefits of Azure Cosmos DB, especially for addressing issues with AI-generated code and documentation, such as inconsistency and unreliability.
Previously, I used minimalistic data formats to ensure consistent outputs and maintain compatibility across different language model versions. The introduction of vector databases improved this by managing large-scale data more effectively, preserving the context and relevance of responses even as models evolved. Exploring vectorization led me to the importance of ontologies, which provide structured frameworks that improve the consistency and relevance of AI-generated content.
I created COSE (Cognitive Optimized Sparse Encoding), a method that uses efficient data formats compatible with vector databases. Integrating COSE enhances data retrieval speeds and computational efficiency through cognitive optimization, making databases more intuitive and adaptive. This synergy boosts performance and propels advancements in AI and data science.
Initially, I started this project to automate my tasks, but it has grown into a solution that I believe will benefit other IT professionals as well.
Link to the Winners' Gallery : https://azurecosmosdb.devpost.com/project-gallery
