The problem Smart Audit solves

Smart Audit is an AI-powered security analysis platform designed to make smart contract auditing faster, more accessible, and more comprehensive.

It addresses several critical pain points for blockchain developers:

• Automates and Accelerates Security Audits : Traditional manual audits are slow and expensive, creating a bottleneck in development. Smart Audit uses multiple AI models to perform a detailed security analysis in minutes, allowing developers to identify and fix vulnerabilities early and often. This makes the task of securing code significantly easier and safer.
• Provides Deeper, Multi-Faceted Insights : Instead of relying on a single perspective, the tool leverages a suite of AI models (from providers like Google, OpenAI, Anthropic, and xAI). This multi-model approach provides a more robust and comprehensive analysis, uncovering a wider range of potential issues than a single tool might.
• Simplifies Gas Optimization : Beyond security, the platform analyzes contracts for gas optimization opportunities. This helps developers write more efficient code, reducing transaction costs for users of the decentralized application.
• Streamlines the Auditing Workflow : The tool provides a user-friendly interface where developers can simply input a contract address from a supported blockchain. The application automatically fetches the source code—even correctly identifying and handling complex proxy contracts—and presents it in a built-in file explorer before running the analysis. This simplifies the entire process from code retrieval to final report.

Link to Slides: https://drive.google.com/file/d/10m77uUOc0Y2wwjXQkG5q4YlEG_GTnEWK/view?usp=sharing

Challenges I ran into

Building Smart Audit involved overcoming several technical hurdles to create a seamless and powerful user experience:

• Integrating Diverse AI Providers : One of the core features is its multi-model analysis. A significant challenge was creating a unified abstraction layer to handle the different API schemas, authentication methods, and response formats for each AI provider (Gemini, OpenAI, Claude, etc.). I had to write specific handlers for each and normalize their outputs so the application could process them consistently.
• Reliably Fetching On-Chain Source Code : Interacting with various blockchain explorers to get verified source code was complex. The biggest hurdle here was correctly handling proxy contracts. I had to implement logic to first detect if an address was a proxy, and then perform additional lookups to find the address of the underlying implementation contract and fetch its source code. This was crucial for providing accurate analysis, and I built the ProxyContractAlert component to make this transparent to the user.
• Managing Asynchronous Operations and State : The entire analysis process is a chain of asynchronous calls: fetching contract info, retrieving source code, and then sending concurrent requests to multiple AI APIs. Managing the application's state through this flow was challenging. I had to carefully handle loading states, errors, and the streaming of AI responses to ensure the UI remained responsive and provided clear feedback to the user without race conditions.
• Crafting Effective AI Prompts and Parsing Reports : The quality of the audit is highly dependent on the quality of the prompts sent to the AI. It took considerable iteration to engineer prompts that could guide the models to perform a nuanced security analysis of Solidity code and identify specific vulnerability categories. Subsequently, parsing the unstructured markdown responses from the AI models and formatting them into a clean, structured, and actionable report in the UI (reportGenerator.ts) was another significant implementation challenge.