products
AI-Powered Data Extraction from Financial Documents
For a UK-based financial services company handling property valuation reports, we created a sophisticated AI system that automatically extracts structured data from unstructured PDF documents.
Key Achievements:
- Processed over 1,000 financial documents
- Maintained complete data privacy with AWS Bedrock
AI Integration Specialist at Visuality
Semantic Search Engine for Knowledge Base
We implemented a sophisticated semantic search system for a client with a large knowledge base. Rather than relying on keyword matching, the system understands the meaning behind search queries, delivering more contextually relevant results.
Key Achievements:
- Implemented vector-based search using PostgreSQL and PGvector
- Created efficient document embedding pipeline
- Optimized for fast search performance even with large document collections
Chief Technology Officer at Visuality
Rapid AI Assistant Implementation for Event Management
For a US-based digital entertainment platform, we rapidly developed an AI assistant prototype that helps users create and set up events through a conversational interface. The assistant collects event information, validates data, and creates the event in the system - all through natural language conversation.
Key Achievements:
- Developed a working MVP in just 3 days with a single developer
- Used prompt engineering to guide the AI assistant's behavior
- Created a feature-flagged implementation for controlled testing
- Proved the viability of AI assistants in streamlining workflows
Ruby & AI Developer at Visuality
Articles
Embeddings
Large Language Models (LLMs) understand human language extremely well. It almost feels like intelligence or magic. However, it's neither. It's a unique algorithm and a set of tools. This article describes one of its core mechanisms and shows how to leverage it using Ruby.
Vector Search
Vector search is a technique used to find items by comparing vectors (embeddings) representing those items. Each item, like a phrase or image, is encoded as a vector in a high-dimensional space, where similar items have vectors that are close to each other.
Vector search is especially useful where keyword or tag-based search methods fall short. Vector Search is great for finding items that represent similar concepts rather than exact string matches.