My Work

Bravous is a sophisticated multi-agent AI platform that fully clones and extends Claude's 'Projects' feature. It provides power users with isolated AI workspaces tailored to specific domains, ensuring strict data separation and focused context. Key technical achievements include: • Multi-Agent Orchestration: Utilizes a LangGraph-powered state graph with specialized agents for planning, researching, and composing. • Dynamic Rendering Engine: A deterministic 3-Gate routing system that outputs standalone files or rich inline UI components (e.g., charts, grids) based on data shape. • Complete BYOK Flexibility: Users can securely integrate their own API keys for OpenAI, local Ollama models, and Tavily Search.

Designed and implemented an autonomous multi-agent research and reporting engine using an orchestrator-worker topology. The system breaks down unstructured, complex user research requests into parallelized sub-tasks, routes them to specialized LLM agents, and enforces strict quality control through an iterative feedback loop.Instead of a rigid linear pipeline, the system utilizes stateful graph routing to dynamically determine the next logical action based on agent outputs, allowing for autonomous self-correction, web-scale data ingestion, and rigorous text editing before final delivery.

Final Year Project: Building a video summarization system using Hugging Face models and implementing a recommender system using Transformer-based architecture for personalized learning experiences.

Context-aware browser extension featuring automated data extraction from active tabs for real-time, webpage-specific AI interaction. It leverages a RAG pipeline to interpret complex concepts directly within the browser.

Built a Few-Shot Consultancy Chatbot using LangChain, FastAPI, and Google Gemini for AI-driven admission guidance and document support. It solves fragmented consultancy through context-aware, multi-mode interactions and serves as a foundation for scalable multi-domain assistants.

Comprehensive machine learning pipeline for predicting real estate prices in Pakistan using Zameen.com dataset. Implemented advanced regression techniques and feature engineering.

Conducted comprehensive exploratory data analysis on global internet usage patterns. Visualized trends and insights using Matplotlib and Seaborn with interactive dashboards.