Custom Model Selected
Response times are typically 30-60 seconds. The model runs on Modal's free tier GPU (T4). First query may take longer due to cold start.
Ask me anything about quantum computing
I can help explain concepts like qubits, superposition, entanglement, quantum gates, and more.
Server sleeps after ~5 minutes of inactivity, so first or idle queries may take 10-15 seconds to wake up.
Try one of these questions:
Available Models
Choose between two inference options based on your needs.
Groq (Llama 3.3 70B)
Response time: 2-3 seconds
Model: Llama 3.3 70B Versatile
Provider: Groq API (free tier)
Best for: Fast, production-quality answers
Recommended for general use. Uses RAG context from 28,071 Q&A pairs.
Custom (QuantumLLM 140M)
Response time: 30-60 seconds
Model: Custom 140M parameter transformer
Provider: Modal GPU (T4, free tier)
Best for: Demonstrating custom ML training
Trained from scratch on OSU's HPC cluster with H100 GPUs. Portfolio demonstration piece.
Both models use the same RAG retrieval system. The custom model showcases end-to-end ML engineering skills, while Groq provides fast, high-quality responses for practical use.
About This Project
A full-stack ML application showcasing end-to-end machine learning engineering,
from custom model training to production deployment.
