A structured curriculum across 13 faculties. Each faculty covers a core discipline of building and operating AI memory systems.
Foundations of persistent AI memory
Explore how AI agents store, retrieve, and forget information. Covering Honcho, Mem0, file-based persistence, semantic search, embedding strategies, and memory compaction.
Crafting effective context windows
Learn to build, budget, and optimize context windows. Covers context builders, budget management, token optimization, sliding windows, and dynamic context scaling.
Understanding LLM internals
Deep dive into transformer architectures, attention mechanisms, MoE layers, quantization, and how different model families (Llama, DeepSeek, Qwen) make trade-offs.
Connecting to any AI backend
Master the provider adapter pattern — integrate OpenAI, Anthropic, Google, local models (Ollama, LM Studio), and custom endpoints through a unified interface.
Building autonomous AI agents
Principles of agent architecture: tool-use, reasoning loops, goal decomposition, multi-agent coordination, and the event-bus pattern for scalable agent meshes.
Seeing inside your AI system
Trace requests across providers, monitor context usage, log agent decisions, set up dashboards, and build alerting for production AI workloads.
Safe and compliant AI operations
Policy engines, PII redaction, access control, encryption at rest and in transit, audit logging, and compliance patterns for enterprise AI deployments.
Making AI fast and efficient
Latency reduction, throughput tuning, caching strategies, batch processing, speculative decoding, and hardware-aware optimization for LLM inference.
Confidence through rigorous testing
Test-driven development for AI prompts, evaluation harnesses, regression testing, hallucination detection, and quality gates for agent outputs.
From laptop to production
Containerization with Docker, orchestration with Kubernetes, CI/CD pipelines, blue-green deployments, and infrastructure-as-code for AI services.
Building elegant AI interfaces
Design principles for AI-facing APIs: consistent schemas, streaming, error handling, rate limiting, versioning, and the single-API gateway pattern.
Powering AI with quality data
Data pipelines for RAG, embedding generation, dataset curation, deduplication, data versioning, and vector database operations.
Responsible AI by design
Fairness auditing, bias detection, transparency reporting, human-in-the-loop patterns, and governance frameworks for responsible AI deployment.
13 faculties · Continually updated · Designed for colorblind safety