docs: add Collabs README explaining Read/Explore/Build learning journey

Author: profvjreddi

collabs/README.md

@@ -0,0 +1,74 @@
+# Collabs
+
+**Understanding the Interplay Between Algorithms and Systems**
+
+> **Status:** Coming Summer 2026
+
+---
+
+## What Are Collabs?
+
+Collabs are hands-on Google Colab simulations that bridge the gap between **reading about ML systems** (the textbook) and **building them from scratch** (TinyTorch).
+
+```
+┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
+│                 │     │                 │     │                 │
+│    Textbook     │────▶│     Collabs     │────▶│    TinyTorch    │
+│                 │     │                 │     │                 │
+│  Concepts &     │     │  Experiment &   │     │  Build from     │
+│  Theory         │     │  Explore        │     │  Scratch        │
+│                 │     │                 │     │                 │
+└─────────────────┘     └─────────────────┘     └─────────────────┘
+      READ                    EXPLORE                  BUILD
+```
+
+## The Learning Journey
+
+| Phase | Resource | What You Do |
+|-------|----------|-------------|
+| **Understand** | [Textbook](https://mlsysbook.ai) | Learn concepts, theory, and system design principles |
+| **Experiment** | Collabs | Explore tradeoffs, tweak parameters, see how decisions ripple through systems |
+| **Build** | [TinyTorch](https://mlsysbook.ai/tinytorch) | Implement everything from scratch, own every line of code |
+
+## Why Collabs?
+
+ML systems are where algorithms meet hardware. A model that works perfectly in theory can fail in practice due to memory limits, latency constraints, or numerical precision. Collabs help you develop intuition for these algorithm-system interactions.
+
+- **See the tradeoffs** — How does batch size affect memory? How does quantization affect accuracy?
+- **Explore interactively** — Adjust parameters and watch how changes ripple through the system
+- **Build intuition** — Understand *why* systems behave the way they do, not just *what* they do
+- **Zero setup** — Run directly in your browser via Google Colab
+
+## Example Topics (Planned)
+
+- **Memory vs. Compute Tradeoffs** — Watch how batch size affects memory footprint and training speed
+- **Quantization Effects** — See accuracy degradation as you reduce precision from FP32 → INT8 → INT4
+- **Attention Visualization** — Explore what transformer attention heads actually learn
+- **Optimization Landscapes** — Navigate loss surfaces with different optimizers
+- **Pruning Strategies** — Compare structured vs. unstructured pruning on real models
+
+## Stay Updated
+
+Collabs are under active development. To be notified when they launch:
+
+- [Subscribe to updates](https://buttondown.email/mlsysbook)
+- [Star the repo](https://github.com/harvard-edge/cs249r_book)
+- [Join discussions](https://github.com/harvard-edge/cs249r_book/discussions)
+
+---
+
+## Related Resources
+
+| Resource | Description |
+|----------|-------------|
+| [Textbook](https://mlsysbook.ai) | ML Systems principles and practices |
+| [TinyTorch](https://mlsysbook.ai/tinytorch) | Build your own ML framework from scratch |
+| [Discussions](https://github.com/harvard-edge/cs249r_book/discussions) | Ask questions, share feedback |
+
+---
+
+<div align="center">
+
+**Read. Explore. Build.** *(Collabs coming soon)*
+
+</div>