Anveshak

• 

  Anveshak

Inspiration Traditional AI assistants rely heavily on centralized infrastructure, which raises concerns around data privacy, ownership, and transparency. We wanted to rethink this paradigm by combining AI with decentralization, enabling users to retain full control over their data. AnveshakAI was inspired by the vision of building a trustless, privacy-first AI assistant that can adapt to any domain without compromising user security.

What it does AnveshakAI is a decentralized AI-powered knowledge assistant built on ICP. It allows users to: Upload documents (PDFs, text, code, images) Automatically process and understand content Ask natural language questions Receive context-aware answers with citations and confidence scores The system effectively becomes a domain-specific expert based on user-provided data, enabling personalized and secure knowledge retrieval.

How we built it We designed AnveshakAI using a modern AI + Web3 stack: Frontend: Next.js for real-time interactive UI Backend: Rust-based ICP canisters for decentralized execution Authentication: Internet Identity for secure login AI Engine: Google Gemini API for intelligent responses Vector Search: HNSW-based embeddings for semantic retrieval Document Processing: Chunking, embeddings, and OCR Deployment: DFX SDK on Internet Computer Protocol This architecture ensures seamless integration of decentralized storage, AI inference, and real-time interaction.

Challenges we ran into Securely integrating AI APIs with decentralized infrastructure Managing API keys without compromising privacy Implementing efficient vector search under blockchain constraints Handling large-scale document processing efficiently Achieving smooth real-time streaming responses Accomplishments that we're proud of Built a fully functional decentralized AI assistant on ICP Achieved sub-2 second response latency Implemented scalable vector search (1M+ embeddings) Enabled multi-format document understanding (including OCR) Delivered a clean, real-time streaming user experience

What we learned How to effectively combine Web3 + AI systems in production-like environments Deep understanding of ICP architecture and canister-based development Practical implementation of semantic search and vector databases Importance of secure key management in decentralized apps Designing systems that balance scalability, latency, and usability

What's next for Anveshak Moving toward fully on-chain AI inference (reducing external API dependency) Advanced domain-specific fine-tuning Multi-user collaboration on shared knowledge bases Mobile-friendly UI and broader accessibility Integration with additional decentralized storage solutions

Built With

• aiml
• icp