CompanyBrain AI

Inspiration

Every company stores its knowledge in scattered places — Google Docs, Slack, Notion, PDFs, spreadsheets, and forgotten folders. Finding the right information wastes hours every week. We wanted to create an AI-powered company memory that never forgets, instantly surfaces answers, and lets teams truly talk to their own data.

What it does

CompanyBrain AI turns all of your company’s data into an intelligent, searchable knowledge base.

• Upload any file — reports, notes, meeting summaries, or datasets.
• The system indexes and understands it using Elasticsearch hybrid (text + vector) search.
• Ask natural questions like “What were our Q2 sales?” or “Summarize feedback from the marketing team.”
• The AI retrieves relevant context from Elastic and uses Gemini AI to generate accurate, cited responses — like having a personal analyst who knows everything your company does.

How we built it

• Frontend: Next.js 15 + Tailwind CSS + Framer Motion for a fast, elegant chat interface and upload dashboard.
• Backend: Node.js API routes handle uploads, parsing, and Elastic indexing.
• Search Engine: Elasticsearch 8.0 with dense_vector fields for embeddings and hybrid BM25 + vector search.
• AI Layer: Gemini 2.0 Flash processes user questions, consumes Elastic results, and produces grounded answers.
• Security: Zod validation, input sanitization, and per-endpoint rate limiting.
• Hosted on Vercel with Elastic Cloud for scalability.

Challenges we ran into

• Configuring Elastic’s vector mappings and hybrid queries correctly took experimentation.
• Managing large document embeddings within query size limits required batching and smart truncation.
• Balancing context length with Gemini’s token constraints while keeping responses coherent.
• Ensuring file parsing worked reliably across multiple formats (PDF, CSV, TXT).

Accomplishments that we're proud of

• Built a fully functional AI knowledge engine that can answer complex company-specific questions in seconds.
• Achieved seamless integration between Elasticsearch and Gemini AI for grounded, context-aware responses.
• Designed a clean, responsive UI that makes enterprise search feel as simple as chatting.
• Demonstrated how Elastic can power intelligent, real-time knowledge systems beyond traditional search.

What we learned

• Deep understanding of Elasticsearch dense vectors and hybrid search techniques.
• How to orchestrate multi-API AI pipelines while maintaining security and scalability.
• The importance of context management and prompt engineering in enterprise AI tools.
• How Elastic’s analytics and aggregations can transform static data into living organizational intelligence.

What's next for CompanyBrain AI

• 🧠 Add multi-user workspaces and role-based permissions.
• 📈 Integrate Elastic dashboards for insights on document coverage and query trends.
• 🌍 Expand to multi-language search using cross-lingual embeddings.
• 💬 Connect directly with Slack, Gmail, and Notion via APIs for continuous knowledge updates.
• ☁️ Launch a hosted SaaS version so any organization can turn its scattered data into one smart, searchable brain.

Built With

• elastic-cloud
• elasticsearch
• framer-motion
• gemini-2.0-flash-api
• javascript
• next.js
• node.js
• react
• tailwind-css
• typescript
• vercel
• zod