Legacy Code Modernization Engine
We leveraged the Fireworks AI API on AMD Hardware because it allowed us to achieve sub-second latency while processing complex, long-context code syntax structures, making real-time codebase refactoring commercially viable.
ReCodeX is an AI-powered enterprise platform built during the AMD Developer Hackathon: ACT II (Track 3 - Unicorn Track). It automatically analyzes aging, "spaghetti" enterprise codebases (such as legacy Java, COBOL, or unoptimized C++), documents their core logic, translates them into modern, secure equivalents, and verifies them for compliance and efficiency.
Instead of a simple chatbot, ReCodeX provides a complete engineering dashboard designed to tackle a multi-billion dollar problem: technical debt and legacy migration.
- Side-by-Side Modernization: Upload a legacy source file and view the original code side-by-side with clean, refactored, modern, and test-covered code.
- Automated Documentation: Deep semantic understanding of code logic to document exactly what the legacy components do.
- Risk & Security Audits: Generates an attached Risk Report highlighting structural changes, security optimizations, and cloud-readiness metrics.
- Developer Pipeline Integration: Built with extensibility to run seamlessly inside automated workflows or CI/CD pipelines.
To satisfy enterprise code processing demands, ReCodeX relies entirely on high-performance cloud infrastructure backed by AMD hardware:
We chose the AMD Developer Cloud because it provided the necessary high-memory headroom and dedicated processing power to handle heavy, parallelized code scanning. Combined with the Fireworks AI API, our architecture achieves sub-second latency while processing complex, long-context code syntax structures.
| Component | Technical Implementation |
|---|---|
| Inference Engine | Core LLM orchestration routed via Fireworks AI API utilizing high-performance AMD hardware endpoints. |
| LLM Core | Powered by Google DeepMind's open-source Gemma models, optimized for deep semantic understanding of complex code bases. |
| Compute Layer | Hosted on the AMD Developer Cloud utilizing powerful cloud-based AMD GPUs to manage parallelized analytics and pipeline tasks. |
| Containerization | Fully containerized architecture using Docker to ensure standard environment reproducibility and effortless deployment. |
This project is fully containerized to ensure uniform performance during evaluation and execution.
- Docker installed locally or on your cloud instance.
- Environment variables set up in an
.envfile (seeexample.envfor reference) containing your Fireworks AI and AMD Developer Cloud credentials.
-
Clone the Repository:
git clone https://github.com/iamharisahsan/ReCodeX.git cd ReCodeX -
Configure Environment Variables:
cp example.env .env
-
Build the Docker Image:
docker build -t recodex-engine . -
Run the Application:
docker run -p 8000:8000 --env-file .env recodex-engine
-
LLM Core: Gemma via the Fireworks AI API (Utilizing the free $50 credit given to hackathon participants).
-
Compute: AMD Developer Cloud (Using the free $100 GPU credits) to host your orchestration layer or do any lightweight prompt caching.
-
Frontend UI: Streamlit & React (Python/Typescript-based, free, and lets you build a polished UI in under an hour).
-
Deployment: Packaged cleanly in a Docker Container. Demo domain deployment on Vercel.
In the landscape of Track 3 (Unicorn Track), ReCodeX focuses heavily on delivering tangible enterprise metrics:
-
Reducing Technical Debt: Decreases ongoing maintenance overhead for mission-critical core logic.
-
Ensuring Regulatory Auditability: Translates undocumented systems into compliant formats with explicit risk mitigation reports.
-
Accelerating Cloud Migration: Transforms monolithic on-premise constructs into cloud-native structures with minimal manual friction.
Lablab.ai | AMD Developer Hackathon: Act-II
Team TechTon