Year

2026

Tech & Technique

Python, FastAPI, React

Description

Built WebScout, an AI-powered research assistant that autonomously answers complex questions by planning, searching, filtering, and synthesizing answers from the web.

Key Features:
  • Intelligent Query Planning - Breaks vague queries into specific, searchable sub-questions
  • Real-Time Web Search - Fetches live data from the web using Tavily API
  • LLM-Based Content Filtering - Uses batch processing to identify relevant sources efficiently
  • Comprehensive Synthesis - Generates well-structured reports with verifiable citations
  • Hallucination Prevention - Grounds responses in real-time web data for accuracy
  • Intent Classification - Differentiates between research queries and small talk to optimize API usage
  • Sequential Search Design - Intentional approach for better relevance and control

Technical Highlights:
  • Backend built with Python and FastAPI for high-performance API endpoints
  • Modern React frontend for seamless user experience
  • Tavily API integration for real-time web search capabilities
  • Pydantic for structured outputs and data validation
  • Stateless architecture for scalability and reliability
  • Batch filtering optimization for efficient source evaluation

My Role

Full-Stack AI Developer
  • Backend: Built FastAPI-based research agent with query planning system
  • Search Integration: Implemented Tavily API for real-time web data retrieval
  • AI Pipeline: Designed LLM-based content filtering and synthesis workflow
  • Intent System: Created intent classifier to differentiate research from chat
  • Frontend: Developed modern React/Next.js interface for research interactions
  • Architecture: Implemented stateless, scalable design with separation of concerns
  • Optimization: Built batch filtering system for efficient source evaluation
WebScout

PADAM