GardaChain
GardaChain is an LLM-based auditor to find vulnerabilities in Smart Contract code
- Microservices
- NextJS
- ExpressJS
- FastAPI
- Go
- Postgresql

About
GardaChain is an LLM-based auditor to find vulnerabilities in Smart Contract code, mainly for Solidity. This is a spin-off project from Bagus Raka's bachelor thesis. GardaChain was originally built for BI-OJK Hackathon 2025, which only earned 5th place, but later made a comeback in Huawei Developer Competition APAC 2025 and earned 2nd place. For the complete background story, read here.
The Problem
Unlike common centralized softwares that can be updated and patched relatively quickly after deployment, Smart Contracts are immutable once deployed. If a bug or vulnerability is found post-deployment, it cannot be updated as easily without workarounds or complete re-deploy of a new contract.
Another problem that was on our radar was that LLMs are harmful to the environment with the resources they use along with the carbon emission they output.
The Solution
The main research under GardaChain was originally to find an approach for LLMs to be more environment friendly without compromising the quality of their output. For the use case, it was aimed at Smart Contracts due to their need of proper auditing before deployment.
GardaChain uses an orchestration of agents to work together. One of them to detect the vulnerabilites, another to find the appropriate fix, another to classify the severity, and so on. With that approach, GardaChain's LLM is much more efficient with only 0.6B and 4B parameters, while still being able to compete with bigger LLMS with 32B parameters. On top of that, GardaChain achieves 98.25% vulnerability detection accuracy.
With these great benefits, we aim to expand GardaChain to cover code beyond Smart Contracts and be an actually profitable business.
