README
AI Engineering - Build and Learn
Welcome to AI Build & Learn, a weekly AI engineering stream where we pick a new topic and learn by building together.
Up next: World Models with DreamerV3 (2026-09-04) An intro to world models: DreamerV3 learns a compact model of its environment and trains a policy inside imagined rollouts. Lightweight and RL-connected. RSVP | Join the Slack
Upcoming:
| 🔨 Topic | 📝 Description |
|---|---|
An intro to world models: DreamerV3 learns a compact model of its environment and trains a policy inside imagined rollouts. Lightweight and RL-connected.
| |
NVIDIA Cosmos, an open world foundation model for physical AI: generative world simulation and action-conditioned rollouts, runnable locally on a DGX Spark.
| |
A different take on world models: V-JEPA 2 predicts in representation space instead of pixels. Self-supervised video understanding and latent planning, runs on a DGX Spark.
|
Past:
| 🔨 Topic | 📝 Description |
|---|---|
Step up from MuJoCo to NVIDIA Isaac Sim: GPU-accelerated, photorealistic robot simulation for physical AI and sim-to-real transfer.
| |
Train reinforcement learning agents to control simulated robots in MuJoCo, the open-source physics engine used as a standard RL benchmark.
| |
Generate music and audio with open-source models: text-to-music and full lyrics-to-song with vocals.
| |
Turn text into speech with open-source TTS: natural voices, voice cloning, and multilingual output.
| |
Open-source text-to-video and image-to-video with diffusion and transformer/DiT models.
| |
Explore the best open-source image models, both diffusion and transformer-based, and try fine-tuning with LoRA.
| |
MLflow for the full ML lifecycle: experiment tracking, model registry & evaluation, LLM tracing, prompt registry, and LLM-as-a-judge, all running on Flyte.
| |
Arize Phoenix is an open-source observability tool for LLM apps: trace, evaluate, and debug your RAG and agent pipelines.
| |
Ragas evaluates RAG and memory systems with structured metrics: retrieval quality, faithfulness, answer relevance, and more.
| |
Cognee is an open-source memory layer for AI agents. Persistent, queryable memory that survives across sessions and tools.
| |
Andrej Karpathy's LLM Wiki concept: build a wiki-style knowledge base specifically structured for LLM consumption and retrieval.
| |
Graph databases with Neo4j for GraphRAG and agent context. When to reach for knowledge graphs, and how to combine graph traversal with semantic search.
| |
Embeddings and vector stores: how semantic search actually works under the hood and how to build RAG pipelines on top.
| |
Build with Google's newly released Gemma 4 models across different sizes for chat, agents, and visual understanding.
| |
Andrej Karpathy's AutoResearch concept: give an AI agent a small but real LLM training setup and let it experiment autonomously overnight.
| |
An e2e framework for creating, deploying and using isolated execution environments for agentic RL training, built using Gymnasium style simple APIs.
| |
Tavily is a search API often used by AI agents. Learn how to integrate Tavily search into your AI applications.
| |
Learn what MCP (Model Context Protocol) is and how to build an MCP server using FastMCP.
|
Setup
# Clone the repository
# Create virtual environment
uv venv .venv --python 3.11
# Activate the venv
source .venv/bin/activate # macOS/Linux
# or
.venv\Scripts\activate # Windows
# Install dependencies
uv pip install -r TopicFOLDER/requirements.txt
Contributors
Similar repos
Languages
Python
94.7%
HTML
2.9%
CSS
1.1%
README
AI Engineering - Build and Learn
Welcome to AI Build & Learn, a weekly AI engineering stream where we pick a new topic and learn by building together.
Up next: World Models with DreamerV3 (2026-09-04) An intro to world models: DreamerV3 learns a compact model of its environment and trains a policy inside imagined rollouts. Lightweight and RL-connected. RSVP | Join the Slack
Upcoming:
| 🔨 Topic | 📝 Description |
|---|---|
An intro to world models: DreamerV3 learns a compact model of its environment and trains a policy inside imagined rollouts. Lightweight and RL-connected.
| |
NVIDIA Cosmos, an open world foundation model for physical AI: generative world simulation and action-conditioned rollouts, runnable locally on a DGX Spark.
| |
A different take on world models: V-JEPA 2 predicts in representation space instead of pixels. Self-supervised video understanding and latent planning, runs on a DGX Spark.
|
Past:
| 🔨 Topic | 📝 Description |
|---|---|
Step up from MuJoCo to NVIDIA Isaac Sim: GPU-accelerated, photorealistic robot simulation for physical AI and sim-to-real transfer.
| |
Train reinforcement learning agents to control simulated robots in MuJoCo, the open-source physics engine used as a standard RL benchmark.
| |
Generate music and audio with open-source models: text-to-music and full lyrics-to-song with vocals.
| |
Turn text into speech with open-source TTS: natural voices, voice cloning, and multilingual output.
| |
Open-source text-to-video and image-to-video with diffusion and transformer/DiT models.
| |
Explore the best open-source image models, both diffusion and transformer-based, and try fine-tuning with LoRA.
| |
MLflow for the full ML lifecycle: experiment tracking, model registry & evaluation, LLM tracing, prompt registry, and LLM-as-a-judge, all running on Flyte.
| |
Arize Phoenix is an open-source observability tool for LLM apps: trace, evaluate, and debug your RAG and agent pipelines.
| |
Ragas evaluates RAG and memory systems with structured metrics: retrieval quality, faithfulness, answer relevance, and more.
| |
Cognee is an open-source memory layer for AI agents. Persistent, queryable memory that survives across sessions and tools.
| |
Andrej Karpathy's LLM Wiki concept: build a wiki-style knowledge base specifically structured for LLM consumption and retrieval.
| |
Graph databases with Neo4j for GraphRAG and agent context. When to reach for knowledge graphs, and how to combine graph traversal with semantic search.
| |
Embeddings and vector stores: how semantic search actually works under the hood and how to build RAG pipelines on top.
| |
Build with Google's newly released Gemma 4 models across different sizes for chat, agents, and visual understanding.
| |
Andrej Karpathy's AutoResearch concept: give an AI agent a small but real LLM training setup and let it experiment autonomously overnight.
| |
An e2e framework for creating, deploying and using isolated execution environments for agentic RL training, built using Gymnasium style simple APIs.
| |
Tavily is a search API often used by AI agents. Learn how to integrate Tavily search into your AI applications.
| |
Learn what MCP (Model Context Protocol) is and how to build an MCP server using FastMCP.
|
Setup
# Clone the repository
# Create virtual environment
uv venv .venv --python 3.11
# Activate the venv
source .venv/bin/activate # macOS/Linux
# or
.venv\Scripts\activate # Windows
# Install dependencies
uv pip install -r TopicFOLDER/requirements.txt
Similar repos
Contributors
Languages
Python
94.7%
HTML
2.9%
CSS
1.1%