Nafizza Azzahra@dorasaki

ModelLens addresses a key challenge with generative AI: AI-generated answers can sound convincing while being unsupported or inconsistent with the original source. This creates a reliability problem when users work with business documents, financial reports, policies, contracts, and other information where answers need to be traceable. ModelLens is an AI document intelligence and verification platform that makes AI answers easier to validate. Users upload a document and ask questions about its contents. The system extracts the document text, divides it into searchable chunks, and retrieves relevant evidence using TF-IDF and cosine similarity. The retrieved evidence is provided to an LLM to generate an answer grounded in the source document. ModelLens then verifies the generated answer against the retrieved evidence. Instead of returning only an AI response, it provides the answer together with a verification status, confidence score, explanation, and source citations showing the relevant evidence and page. The system identifies four outcomes: **SUPPORTED** when the answer is supported by the source, **CONTRADICTED** when it conflicts with the source, **NOT_FOUND** when relevant information cannot be found, and **UNCERTAIN** when the available evidence is incomplete or ambiguous. The target users include analysts, operations teams, compliance users, researchers, and organizations that need reliable answers from document-based information. The interaction is simple: upload a document, ask a question, receive an AI-generated answer, and inspect the evidence used to support or challenge it. What makes ModelLens unique is its verification layer. Traditional document Q&A focuses on retrieving information and generating an answer. ModelLens adds a separate step that evaluates whether the generated answer is actually supported by the retrieved source evidence. This creates a transparent evidence trail and helps users identify unsupported or contradictory responses instead of blindly trusting AI output. ModelLens supports PDF, DOCX, TXT, Markdown, CSV, JSON, and source-code documents. Its implementation combines a React and TypeScript frontend, Python FastAPI backend, PyMuPDF and python-docx for document processing, Scikit-Learn for retrieval, and Groq-powered LLM inference.

💻 Online Hackathon | ⏳ Build: 48 hours | 📅 September 25–27, 2026 The official IBM Bob 2.0 hackathon. Build what's next in AI-assisted development, alongside the developers and experts shaping it. ⭐ Get help from experts throughout the event 🤝 Join a team or build solo 💰 Prize pool: $12,000 🧑🏻‍💻 Register before the Kick-Off Stream to start building with Bob

Build solutions that improve how software is built. Work with an AI that understands your codebase, helping you reduce repetitive work and build with real context. ⏳ Build your project in 48 hours. 🚀 Get hands-on with IBM Bob and explore how AI fits into real development workflows before it becomes standard. 💡 Build tools and workflows that developers would actually use. ⭐ Learn directly from experts throughout the event. 🤝 Join a team or build solo. 💰 Prize pool: $10,000 🧑🏻‍💻 Register before the Kick-Off Stream and start building with Bob.