Featured Projects
Personal AI projects showcasing expertise in LLMs, RAG, and agentic architectures
FinAgent - Autonomous Financial Analysis Agent
AI-powered financial analyst leveraging the ReAct pattern for autonomous tool selection and real-time analysis.
FinAgent is an autonomous financial analyst that uses the ReAct (Reasoning + Acting) pattern to answer investment and financial queries. It autonomously selects appropriate tools, executes them, and delivers transparent, evidence-based responses. The system handles complex queries requiring multiple data sources through natural language interaction, providing real-time financial insights.
Key Features
- •Autonomous tool selection using ReAct pattern (Reasoning + Acting)
- •Real-time streaming with Server-Sent Events (SSE)
- •Comprehensive evaluation: tool selection accuracy, parameter extraction, response grounding
- •Multi-step query handling with transparent reasoning process
- •Stock price lookup (historical and current)
- •Company information retrieval
- •Financial ratio calculations (P/E, ROE, ROA, margins)
- •Investment return projections
Highlights
- ✓Demonstrates agentic architecture with autonomous decision-making
- ✓Real-time reasoning transparency for explainable AI
- ✓Production-ready evaluation metrics for quality assurance
- ✓Full-stack implementation with modern tech stack
Technology Stack
JudRag - Legal Document RAG System
Retrieval-augmented generation system for semantic search over legal documents and case law.
JudRag is a specialized RAG (Retrieval-Augmented Generation) system designed for legal documents. It enables semantic search over judicial documents, case law, and statutes using vector embeddings and conversational interfaces. The system preprocesses legal documents into chunked text, stores them in a vector database, and provides conversational search capabilities for legal professionals.
Key Features
- •Semantic search over legal documents using vector embeddings
- •RAG pipeline with Chroma vector database
- •Conversational interface for case law queries
- •Document chunking and preprocessing for optimal retrieval
- •Session management for conversation history
- •Full-text search combined with semantic similarity
Highlights
- ✓Demonstrates RAG expertise with domain-specific application
- ✓Combines traditional search with modern LLM capabilities
- ✓Full-stack implementation from data processing to UI
- ✓Specialized for legal/judicial domain knowledge
Technology Stack
Note: Flagship AI products developed at Strategy (Agents, Auto Dashboards, Auto Expert) are under company contract and cannot be publicly showcased. The projects above represent personal work demonstrating similar technical capabilities.