👋 Hi, I’m Shravan Reddy

Hackathon Projects

  1. AMD AI Sprint Hackathon 2025 - 4th Place Participated in the AMD AI Sprint Hackathon 2025 at the Indian Institute of Science (IISc), sponsored by Hugging Face and Meta. Our team secured 4th place by building a Question Agent and Answer Agent using Qwen3-4B for an IPL-style game. Explore the Repository

  2. LLM-Kernel-Agent (Cerebras Inference API Agent) A vibe coded project that combines LLMs with kernel functions, leveraging the Cerebras Inference API with the model: llama-4-scout-17b-16e-instruct in Cerebras x Cline during the Vibe Coder Hackathon. This project builds upon the AMD Hackathon project, using the same template and incorporating improvements. The agent accepts high-level natural language prompts, generates optimized Triton GPU kernel code, uses Cerebras as the inference backend, provides a Streamlit UI with secure API key input, and handles UTF-8 content. View Streamlit App Explore the Repository View Video

My Recent Projects

  1. Trained a Llama 8B Model on Custom Dataset
    Implemented in Google Colab.
    View Project

  2. AMD GPU Kernel Optimization Challenge – My Experience
    In April 2025, I participated in AMD’s FP8 GEMM Kernel Optimization Challenge. The task involved writing high-performance matrix multiplication kernels for MI300X GPUs using the Triton framework, targeting FP8 inputs (E4M3FNUZ) and BF16 outputs with per-group scaling.
    View Documentation

  3. Deep Learning Repository
    This repository contains various resources, implementations, and discussions related to deep learning concepts, including neural networks, optimization techniques, and large language models.
    Explore the Repository

  4. Data Structures and Algorithms Repository
    A comprehensive collection of data structures and algorithms implementations in various programming languages, including solutions to common coding problems and interview questions.
    Explore the Repository