H M Sajjad Hossain

InvestAssist: AI-Powered Financial Analysis Platform

Project Overview

Traditional financial analysis tools often provide fragmented insights, requiring users to navigate multiple platforms and manually synthesize information from various sources. Investors and analysts need a unified platform that can:

Provide comprehensive market analysis across multiple dimensions
Deliver personalized insights based on individual portfolios
Process and interpret complex financial data in real-time
Offer strategic recommendations backed by quantitative analysis

InvestAssist is a sophisticated financial analysis platform that leverages cutting-edge AI technology to provide comprehensive market insights and investment recommendations. Built with an innovative agentic architecture using CrewAI, the platform employs specialized AI agents that collaborate to deliver real-time, actionable financial intelligence.

GitHub: InvestAssist: Your AI-Powered Financial Co-Pilot

Demo: InvestAssist Demo

RagFuse

Project Overview

RAGFuse is a sophisticated Retrieval-Augmented Generation (RAG) application that enables users to have intelligent conversations with their documents. Built with Flask and powered by advanced vector search technology, the platform transforms static documents into interactive knowledge bases through natural language querying. (I built this for my wife to run locally since at her work she has some sensitive data that she cannot share)

Core Capabilities

Comprehensive Document Management: Seamlessly upload, organize, and manage PDF, DOCX, PPTX, TXT, MD, CSV, and JSON files with intelligent parsing. Features include customizable folder structures for document organization, built-in PDF viewer for document preview, and easy document deletion and reorganization capabilities.

Advanced RAG Architecture: Utilizes ChromaDB for high-performance vector storage and implements neural-based chunking through the chonkie library's NeuralChunker for intelligent document segmentation, ensuring optimal context retrieval and response quality.

Multi-LLM Integration: Supports multiple AI providers including OpenAI, Claude, Gemini, and Ollama with persistent configuration management, allowing users to choose their preferred language model for document interactions.

Intelligent Chat Interface: Provides context-aware conversations with uploaded documents, featuring optional web search integration for enhanced knowledge augmentation beyond the document corpus.

Technical Architecture

Key Technical Features:

Real-time document processing and indexing
Vector-based semantic search for precise context retrieval
Persistent settings and conversation history
Built-in PDF viewer for document preview
Scalable architecture supporting large document collections

Github: RagFuse: Manage Documents and Talk to Your Documents

TeamBidder

Project Overview

I could not find a simple auction organizing platform for local friendly tournaments among friends. Hence decided to build one for our Seattle soccer group. TeamBidder is a real-time auction system built with Flask and WebSockets that enables live bidding experiences with multi-user support, countdown timers, and comprehensive testing infrastructure. The platform provides an engaging auction environment for team-based bidding scenarios, particularly designed for sports player auctions.

Core Features

Real-Time Bidding Engine: WebSocket-powered live bidding system with instant updates across all connected users, ensuring synchronized auction experiences without page refreshes.

Multi-User Role Management: Secure access control with 8-character invitation codes supporting three distinct roles - Admin (auction control), Bidders (team owners), and Visitors (observers) - each with appropriate permissions and interfaces.

Dynamic Auction Flow: Automated 60-second countdown timer for auction start, followed by timed bidding windows for each player, creating an authentic auction atmosphere with built-in urgency and excitement.

Budget & Team Management: Real-time budget tracking and spending calculations for each bidding team, with configurable maximum players per team and total budget constraints to ensure fair competition.

Interactive Communication: Live chat functionality enabling real-time communication between participants during auctions, fostering engagement and competitive dynamics.

GitHub: sajjadriaj/teambidder

FootMob

Project Overview

A Flask-powered web application that transforms how small football groups organize matches. Features three AI-driven team balancing algorithms (Smart Draft, Multi-Armed Bandit, and Simulated Annealing) that analyze player skills, historical performance, and recent form to create perfectly balanced teams.

GitHub: https://github.com/sajjadriaj/footmob

A version is deployed here:

Website: FootMob Web demo