Price: 0
Number of applications: 14
10.07.26 (inclusive)
Взаимодействие на партнерских началах
Idea
ICT tasks
Media sphere
Other technological solutions
Software/ IS
In the activities of modern IT and service companies, employees are daily confronted with a huge array of internal information: technical regulations, instructions, project documentation, guidelines and legal agreements. Manually searching for the necessary data and answers to specific questions inside corporate Wiki systems and network folders takes a significant amount of work time and reduces the overall effectiveness of teams.
The implementation of the task will automate the search for corporate documents, reducing the time for employees to obtain the necessary information by 3-5 times. By deploying the system strictly in the local loop (On-Premise), the risks of data leakage and violations of the NDA will be completely eliminated. The company will receive an independent AI tool based on Open-Source technologies that works without paid foreign subscriptions and external APIs. This will reduce the burden on relevant departments due to automatic answers to standard questions in Kazakh and Russian.
Gabdulov D.J.
Purpose and description of task (project)
Exploring the possibilities and creating a prototype (Proof of Concept) of a local corporate AI assistant based on Open-Source language models (LLM) with RAG architecture, capable of quickly and securely performing contextual search and analysis of information on internal regulations and company documents without transferring data to external cloud services. Task description: As part of the technological task, it is proposed to solve the problem of safe and fast information search inside corporate databases. Modern public AI solutions (ChatGPT, Claude, etc.) cannot be used in the company's circuit due to strict information security requirements and NDA conditions. The task is to deploy and adapt an open language model (for example, Llama 3, Mistral or analogues) completely on local capacities. The project includes: Creating a module for downloading and indexing text documents (PDF, DOCX, TXT); Configuring a vector database for precise context extraction (RAG); Development of a user-friendly web interface (chatbot) that will answer questions from employees in Kazakh and Russian, necessarily referring to specific paragraphs and pages of primary source documents. The result of the task should be a workable prototype packaged in a Docker container for demonstration to the internal technical team.
Whatsapp
