ANA :Agentic Corrective RAG using LangGraph

Leverages LangGraph for modular workflow design and node-based task orchestration. The system integrates LLM with document retrieval and web search to mitigate hallucinations and improve response accuracy by dynamically fetching relevant context. Designed as an agentic architecture, it incorporates corrective mechanisms for enhanced reliability in query-answering workflows.

Demo | Code

LLM from scratch

Implements a GPT-like language model from scratch, focusing on core components like tokenisation, multi-head attention, feedforward layers, and decoder blocks. The model is trained to generate text by predicting the next word in a sequence using a decoder-based transformer architecture.

Blog | Code

Financial Assistant

Provides insights into key financial metrics like liquidity, solvency, and profitability for US-SEC listed companies. Leveraging LangChain's tool-calling agent, and data visualization, it dynamically generates accurate responses and interactive plots based on user queries. This solution bridges advanced NLP capabilities with financial analytics, offering a robust tool for professionals and enthusiasts alike.

Demo | Code

Spendwise

A personal financial assistant utilizing data analytics techniques for transaction categorization, spending insights, and chat query support.

Demo | Code

Multi Agent

Leverages a robust multi-agent system to generate detailed project plans tailored to user inputs. It features dynamic project planning, employing a multi-agent architecture for task decomposition, resource estimation, and effective allocation.

Demo | Code