Finding One Service Provider for Ai Platform Development Tender

• Country: Czech Republic
• Summary: Finding One Service Provider for Ai Platform Development
• CZT Ref No: 140016804
• Deadline: 03 May 2026
• Financier: Self Financed
• Purchaser Ownership: Government
• Tender Value: Refer Document
• Notice Type: Tender
• Document Ref. No.: N006/26/V00011788
• Purchaser's Detail:
  Login to see full purchaser details. Login to see full purchaser details. Login to see full purchaser details. Login to see full purchaser details. Login to see full purchaser details. Login to see full purchaser details. Login to see full purchaser details. Login to see full purchaser details. Login to see full purchaser details.

  Login to see details
• Description:
• The five requests include, in particular: GPU Infrastructure and Model Serving: An operational lightweight GPU instance (T4/L4) for the local guest model. Implementation of the vLLM tool for parallel handling of the embedding model (multilingual-e5-large) and large language models (Gemma 3 12B / Llama 3.1 8B). Development of REST API for double and individual generation embedding with mechanisms of health checks and automatic clowning. Vector search engine and DB layer: Implementation of pgvector in PostgreSQL for storing and browsing vector representations of data. Development of hybrid a compact system combining cosine similarity with hard filters Architectural preparation for future migration to a dedicated Qdrant vector database. AI Scraping Engine and Normalization: Integration with the Apify platform for collecting dozens from the Internet Vyuit self-hosted LLM for entity extraction and interpretation (CZ-EN) within one inferential framework using JSON mode and Zod schmat for guaranteed input structure. Implementation of deduplicated Gatekeeper mechanism using SimHash/MinHash algorithms for fingerprinting advertisements. Pipeline for processing of embedded pdf data: A system for extracting text from PDF/DOCX formats with an implemented OCR fallback. Automated detection of language and structured unstructured data from a document into a defined JSON object using local LLM.11 Logic for automatic pepoet embedding of structured data when data changes and security limits for update frequency. AI Matching and Reasoning: The overall model of agreement (multi-score model) including the agreement of those groups of information. Generating text derived matches through external APIs (OpenAI / Anthropic). The different categories of reasoning. A complex system for caching the result in Redis using SHA256 hashes (User_ID + Item_ID + Profile_Hash) with a 30-day validity. Asynchronous processing and orchestration: Use Redis and BullMQ for managing and prioritizing asynchrono...
• Documents:

 Tender Notice