Inspiration

Performance optimization is often difficult because developers need to manually find CPU bottlenecks, understand hardware capabilities, modify code, and then prove that the optimization actually works.

We built Performix to make this workflow safer and more measurable. It analyzes Python projects, identifies optimization opportunities, generates an optimization strategy, requires human approval, executes the approved transformation, and validates the result.

What it does

Performix is an AI-powered performance engineering and validation platform designed with ARM CPU readiness in mind.

The workflow is:

Project Scan → Performance Analysis → AI Optimization Plan → Human Approval → Safe Execution → Correctness Validation → Benchmarking → Executive Report

It detects CPU-intensive patterns such as nested loops and matrix multiplication, recommends vectorization strategies, performs deterministic optimization when the local LLM is unavailable, and measures runtime and memory changes.

How we built it

Performix is built with Python and uses AST-based source analysis, deterministic optimization logic, subprocess benchmarking, memory measurement, local GGUF/LLM support, and a desktop GUI.

The platform uses a safety-first workflow where the original project is backed up before execution. Optimization is not applied until human approval is received.

Challenges we faced

One of the biggest challenges was making benchmarking reliable across different Python projects. We also had to handle packaged PyInstaller execution, dynamic model paths, custom function names, and projects that do not follow a fixed folder structure.

We solved these issues by adding dynamic target discovery, wall-clock benchmark fallbacks, AST-based pattern detection, dynamic model-path resolution, and deterministic optimization fallbacks.

Results

In our final test project, Performix successfully detected a matrix multiplication optimization opportunity, applied the deterministic optimization path, validated correctness, and measured a 336.83× calculated runtime improvement.

The execution pipeline also generated JSON and HTML executive reports containing the benchmark and validation results.

Why it matters

Performix turns performance optimization from a manual trial-and-error process into a controlled engineering workflow with analysis, approval, safe execution, validation, and measurable results.

Built With

Share this project:

Updates