A fully voice-driven System Design RAG (Retrieval-Augmented Generation) pipeline in Hindi — powered by Whisper, Groq, Pinecone, LLaMA, and SpeechT5 built as a side project.
ASR Whisper Tiny (finetuned) Finetuned for 2000 epochs (https://discuss.huggingface.co/t/whisper-model-fine-tuning/26045)
Translation Groq API For fast Hindi-to-English conversion
Vector DB Pinecone 6500+ curated Q&A vectors
LLM LLaMA 3.3 70B Versatile Generates context-aware responses
TTS SpeechT5 (finetuned) Trained for 1500 epochs on English speech
To avoid session timeout during long model training in Google Colab, use the following JavaScript snippet in your browser's developer console:
// 🔁 Auto-reconnect for Google Colab
function ConnectButton() {
console.log("Connect pushed");
document.querySelector("#top-toolbar > colab-connect-button")
.shadowRoot.querySelector("#connect").click();
}
setInterval(ConnectButton, 60000);