hackjakarta

New Version: https://github.com/muazhari/autocode

Team

• Muhammad Kharisma Azhari (Stackup: muazhari)
• Vincent Yono (Stackup: vincentyono)

Idea

Challenge Statement

Digital Empowerment

In alignment with Grab’s mission to “Drive Southeast Asia forward by creating economic empowerment for everyone”, your challenge is to harness Generative AI (text, image, or video) to innovate and elevate Grab's Passenger, Driver, or Merchant mobile application’s core journey (e.g., transport, food, mart). Develop a creative, scalable solution using Generative AI that improves user experience within the Grab app, drives economic growth, and empowers individuals and businesses within Southeast Asia.

Proposed Project

Auto Code Improvement by User Experience and Technical Metrics Optimization

We propose to improve the user experience and technical metrics of Grab applications. Specifically, we improve user experience and technical metrics using generative AI. Based on our research/literature review, our project hypothetically can contribute to the user experience and economic performance of the company.

Project Scope*

• User Experience Metrics: Error Potentiality, Latency.
• Technical Metrics: Code Quality.

*Can be extended to other metrics, like throughput.

Project Future Roadmap

• Direct frontend evaluation using reinforcement learning as real-user simulator. Metrics measured by how easy the agent "wants" to be fulfilled.
• Auto system architecture search. Inspired by neural architecture search.

Tech Stack

• Python
• Golang
• Langchain
• Pymoo
• OpenAI

Usage

1. Clone the repository
2. Change directory to ./client/app_product
3. Run go mod tidy to install dependencies.
4. Change directory to ./server/autocode
5. Run pip install . to install dependencies.
6. Run cell for optimization instantiation in ./server/autocode/example.ipynb
7. Run go test ./test in app_product working directory.
8. Run cell for optimization.run() in ./server/autocode/example.ipynb to start the optimization process.
9. Open dashboard in http://localhost:{dashboard_port} to see the optimization process in real-time.
10. Wait until the optimization process is finished.
11. Decide the best solution from the optimization process result.