• 

  Landing Page

QuickMafs: LLM-integrated Code Generator for Embedded Convex Optimization

What is QuickMafs

QuickMafs is a LLM-integrated code generator for embedded convex optimization, addressing the challenges engineers face in translating complex mathematical expressions with convex objectives into efficient C code for embedded devices. It ultimately facilitates applications ranging from aircraft flight control systems to autonomous vehicles, portfolio optimization, healthcare systems, and smart grids.

Problem Overview:

What is Convex Optimization?

• Potent tool with diverse applications from finance to healthcare and control systems, empowers engineers to find optimal solutions in complex scenarios.

Why is Convex Optimization in Embedded Devices Important?

Aircraft flight control system

• Maintaining stability during atmospheric disturbances,

Autonomous Vehicles:

• Path Planning, Trajectory Optimization, Collision Avoidance
  

Portfolio Optimization:

• Portfolio Optimization, Risk Management, Algorithmic Trading Strategies

Healthcare Systems:

• Patient Scheduling, Resource Allocation in hospitals, Optimization of medical treatments

Smart Grids:

• Energy Distribution, Minimize transmission losses, schedule power generation efficiently, load balancing.
  

Problem faced by Engineer?

• The translation of intricate mathematical expressions with convex objectives into efficient C code for embedded devices is time-consuming and prone to errors!

Why use our QuickMafs

• It is a code generation flow that not only optimizes the conversion but also resonates with the practical needs of engineers.

What are the features of QuickMafs

• Image-to-Mathematical-Expression Conversion
  
• Large Language Model with Prompt Engineering
  
• Export generated code as .c or .txt
  

System Architecture

The system features a React frontend communicating with an Express.js server. TroCR-based-printed models handle image-to-math expression conversion, while GPT-4 models classify problem statements and convert expressions/high-level code to C. Output data is stored in a local DB for convenient future retrievals.

Code Generation Flow

1. Math Prompt

Input your math prompt in plain English expression or in the supported programming languages.

2. Generate

Our system will process your input and return with the equivalent embedded system code

3. Save Result

Immediately use the generated code by either copying the code directly to clipboard or saving as file.

Built With

• azure
• express.js
• hugging-face
• large-language-model
• node.js
• openai
• react
• transformer