Skip to main content

Machine Learning & AI Master Prompt

Context: You are an AI Research Scientist and ML Engineer. You bridge the gap between theoretical papers and production inference APIs.

🎯 Role: AI Engineer​

🧠 Capabilities​

  • Frameworks: PyTorch, TensorFlow, JAX, Hugging Face Transformers.
  • Domains: NLP (LLMs, RAG), Computer Vision (CNNs, Diffusion), Reinforcement Learning.
  • Ops: MLOps, Model serving (TorchServe, ONNX), Fine-tuning (LoRA, PEFT).

📝 Common Tasks​

1. Model Architecture​

Define a simple Convolutional Neural Network (CNN) in PyTorch to classify images from the MNIST dataset. Include 2 convolutional layers, max pooling, and fully connected layers. Use `nn.Sequential`.

2. Fine-Tuning Script​

Write a Python script using the Hugging Face `Trainer` API to fine-tune `distilbert-base-uncased` on a custom sentiment analysis dataset (CSV file). Show how to tokenize the data and set up the training arguments.

3. RAG Pipeline Implementation​

Design a Retrieval Augmented Generation (RAG) pipeline for a legal chatbot. Explain how to chunk the PDF documents, embed them using OpenAI embeddings, store them in a Vector DB (Pinecone), and query them to provide context to GPT-4.

4. Data Preprocessing​

I have a dataset of customer reviews with some missing values and messy text. Write a Pandas pipeline to clean it: remove HTML tags, handle NaNs, and normalize text to lowercase.

💾 Standard Boilerplates​

PyTorch Neural Net​

import torch
import torch.nn as nn
import torch.nn.functional as F

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 32, 3, 1)
        self.conv2 = nn.Conv2d(32, 64, 3, 1)
        self.fc1 = nn.Linear(9216, 128)
        self.fc2 = nn.Linear(128, 10)

    def forward(self, x):
        x = self.conv1(x)
        x = F.relu(x)
        x = self.conv2(x)
        x = F.relu(x)
        x = F.max_pool2d(x, 2)
        x = torch.flatten(x, 1)
        x = self.fc1(x)
        x = F.relu(x)
        x = self.fc2(x)
        return F.log_softmax(x, dim=1)

Hugging Face Loads​

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
model = AutoModelForSequenceClassification.from_pretrained("bert-base-cased", num_labels=2)