Skip to content
This repository was archived by the owner on Feb 7, 2026. It is now read-only.

ARUNAGIRINATHAN-K/gpu-accelerated-ds-agents

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

101 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Autonomous data science workflows with 10-50× GPU speedup

Python RAPIDS PyTorch License


A production-ready system for autonomous data science leveraging NVIDIA CUDA and RAPIDS on Google Colab's free GPUs. Run complete ML pipelines from EDA to deployment with zero API costs using local LLM orchestration.


Features

Autonomous Agents

  • EDA Agent - Auto data profiling
  • Modeling Agent - Multi-algorithm training
  • Viz Agent - Smart visualizations
  • Report Agent - Full documentation

GPU Acceleration

  • 10-50× faster than CPU
  • Free Tesla T4/V100 on Colab
  • RAPIDS cuDF & cuML
  • XGBoost GPU support

Local LLM Orchestrator

  • Zero API costs
  • Llama 2 / Mistral / Phi
  • Intelligent workflow planning
  • Context-aware decisions

Hybrid Architecture

  • Local Streamlit UI
  • Cloud GPU execution
  • Seamless data transfer
  • Flexible deployment

Tech Stack

Category Technologies
GPU Computing CUDA RAPIDS
ML Frameworks PyTorch XGBoost cuML
LLM Stack Transformers Llama
Data Processing cuDF Pandas
Visualization Plotly Matplotlib
UI/Platform Streamlit Colab

Benchmarks

Operation CPU Time GPU Time Speedup
Load 10M rows 45s 2.3s 19.6×
Preprocessing 120s 6.5s 18.5×
Correlation 35s 1.2s 29.2×
XGBoost 180s 12s 15.0×
Full Pipeline 380s 22s 17.3×

Tesla T4 GPU vs Intel i7-10700K CPU


Built with ❤️ for the data science community

GitHub LinkedIn