📅 Day 1: Autonomous Bio-Research – Building Smart AI Literature Engines
- Core Objective: Build an AI-powered biomedical literature engine that can retrieve, search, and answer research questions from scientific documents.
- Architecture of a modern, domain-specific Bio-RAG pipeline
- Semantic chunking for parsing complex biomedical and medical text
- Creating deep-learning searchable vector indices from scientific knowledge
- Prompt guardrails to reduce hallucination and improve research reliability
🛠️ Hands-on:
- Hands-on Lab: Build and deploy a local Literature QA Bot in Google Colab
🧰 Tools Covered: Bio-RAG, Vector Indexing, Prompt Engineering, Google Colab
📅 Day 2: AI-Driven Design – Generating Synthetic Biological Sequences with PyTorch
- Core Objective: Learn how deep learning models can generate synthetic biological sequences from sequential DNA data.
- Deep learning for DNA sequence modeling using LSTMs
- Processing large FASTA files using PyTorch data handling workflows
- Setting up, optimizing, and training a neural network from scratch
- Autoregressive inference for generating new biological sequences
🛠️ Hands-on:
- Hands-on Lab: Train an AI sequence generator and observe real-time synthetic biological sequence generation
🧰 Tools Covered: PyTorch, LSTM Models, FASTA Files, Google Colab
📅 Day 3: Protein Transformers & Attention Mapping
- Core Objective: Understand how protein transformer models learn molecular patterns and visualize attention signals in structural biology.
- Transformer architectures in the context of molecular biology
- Feature extraction using Meta’s ESM-2 protein language model
- Slicing transformer layers and mapping attention arrays
- Translating attention matrices into structural alignments and biological insights
🛠️ Hands-on:
- Hands-on Lab: Map and visualize protein transformer attention on a 3D structural model
🧰 Tools Covered: ESM-2, Protein Transformers, Attention Mapping, 3D Structural Visualization
