Looking for an AI solution for your own information workflow? Contact OBLIQUID to discuss document intelligence, semantic search, RAG and custom NLP applications.
turning documents into answers
AI and natural language processing, engineered for real-world use
FEP is an AI platform fully developed in-house by OBLIQUID to collect, understand and search large collections of documents.
It brings together automated information gathering, natural language processing (NLP) and conversational AI. Instead of reading through scattered sources, users can find relevant information, explore a topic and ask questions in everyday language.
Automatically acquire documents from selected web sources, clean their text and bring them into a searchable knowledge base.
Translate and summarize texts, classify them by topic and extract entities, relationships and numerical data. Turn unstructured content into organized information.
Combine keyword search, topic filters and semantic search, which finds relevant content by meaning rather than exact wording. Ask questions through a conversational interface.
Not just a chatbot: answers grounded in your sources.
The most demanding part of FEP was engineering its Retrieval-Augmented Generation (RAG) pipeline. Inspired by Perplexity.ai, it implements the same core research experience: ask a question, retrieve relevant information and receive a synthesized answer with references to the sources.
FEP searches the document collection for information relevant to the question, using semantic retrieval to go beyond keyword matches.
The retrieved content is supplied to a large language model (LLM) as context, so the answer is based on the selected documents rather than only on the model's general knowledge.
Users receive a readable synthesis, can inspect its sources and continue with follow-up questions to investigate the topic in more detail.
This makes research easier to check and helps reduce unsupported answers. Source references remain essential: AI-generated answers still require appropriate human review.
This is implemented software, not a concept or a specification. FEP connects document acquisition, AI processing, retrieval and answer generation in one working platform.
A production application and further ways businesses can put FEP's capabilities to work:
Make a company document collection easier to consult. Sales and support teams can ask about product features, procedures or technical guidance and return to the original material for verification.
Example: "Which product meets these requirements, and what documentation supports it?"
Follow selected news and specialist sources, organize updates by topic or company and summarize developments without manually reading every article.
Example: "What are these sources reporting about new uses of this material?"
FEP's RAG pipeline has been used in production for environmental, social and governance (ESG) assessment: retrieve documentary evidence, extract numerical values and generate structured answers and scores for sector-specific questions.
The same approach can support regulatory research, helping specialists find and assess relevant evidence without replacing expert judgment.
Translate and summarize international sources, search across a topic and generate focused reports. Give different teams access to information relevant to their interests and responsibilities.
Example: "Prepare a briefing on this topic using the documents in our collection."
FEP uses a scalable microservices architecture: separate services handle crawling, AI analysis, orchestration, storage and web interfaces. Processing resources can scale with the workload.
Its technology includes Python for NLP, Node.js and React for application services and interfaces, Elasticsearch for document search and Qdrant for vector-based semantic retrieval, alongside open-weight language models.
Administrators manage sources, topics and access permissions. AI-enriched documents can be reviewed and validated by an operator, while dedicated web portals expose the appropriate content to each audience.
OBLIQUID developed the entire platform, from data acquisition to the conversational user experience. This work demonstrates our ability to engineer and integrate complex AI systems, not simply connect a chat interface to a model.