The map
One graph. Every concept. A clear path.
Every concept is a node. Every edge is a prerequisite. Tracks run left to right, week by week, follow an edge backward and you find exactly what to master first. No guessing what to learn next.
Concept map
Mapped end to end.
Curved lines connect a concept to its prerequisites. The visual is decorative; the full, screen-reader-friendly version follows as a table.
Agentic Systems & Harness EngineeringAI Infrastructure & InferenceApplied ML & Model EngineeringProduction AI ProductsFrontier SystemsMultimodal AI & Embodied World ModelingAI Safety, Alignment & InterpretabilityLand the Elite AI Role
Every concept, in order
The graph as a table.
A complete, accessible listing of every node, the classic CS subject it bridges to, and its prerequisite edges.
| Week | Track | Concept | Bridges to | Prerequisites |
|---|---|---|---|---|
| 1 | Agentic Systems & Harness Engineering | Foundations of Harness Engineering & Loop Crafting | Operating Systems — processes, scheduling, and the run loop | None |
| 1 | AI Infrastructure & Inference | Transformer Internals for Serving | Computer Architecture — instruction-level parallelism and the memory wall | None |
| 1 | AI Safety, Alignment & Interpretability | The Alignment Problem & Safety Foundations | Theory of Computation — specification, correctness, and the limits of formal goals | None |
| 1 | Applied ML & Model Engineering | Neural Networks & Backpropagation from Scratch | Calculus & Linear Algebra — gradients, the chain rule, and vector spaces | None |
| 1 | Frontier Systems | Distributed Systems Foundations | Distributed Systems — failure models, latency, and the CAP theorem | None |
| 1 | Land the Elite AI Role | AI / ML System-Design Interviews | , | None |
| 1 | Multimodal AI & Embodied World Modeling | Vision Encoders & Image Representation | Computer Vision — image representation, feature extraction, and convolution | None |
| 1 | Production AI Products | LLM Application Foundations | Software Engineering — application architecture and API design | None |
| 2 | Agentic Systems & Harness Engineering | Dynamic Inline Tooling, MCP & Cache-Preserving Toolmaking | Software Engineering — interface design and API contracts | Foundations of Harness Engineering & Loop Crafting |
| 2 | AI Infrastructure & Inference | GPU Architecture & the Memory Wall | Computer Architecture — parallelism, memory hierarchy, and the roofline model | Transformer Internals for Serving |
| 2 | AI Safety, Alignment & Interpretability | RLHF, Direct Alignment & Verifiable Reasoning (GRPO) | Machine Learning — reinforcement learning and optimization under feedback | The Alignment Problem & Safety Foundations |
| 2 | Applied ML & Model Engineering | Deep Learning Foundations | Machine Learning — optimization, generalization, and the bias-variance tradeoff | Neural Networks & Backpropagation from Scratch |
| 2 | Land the Elite AI Role | Running Experiments, Ablations & Tracking | , | AI / ML System-Design Interviews |
| 2 | Multimodal AI & Embodied World Modeling | Vision-Language Model Architecture | Machine Learning — representation learning and modality fusion | Vision Encoders & Image Representation |
| 2 | Production AI Products | Embeddings & Vector Search | Databases — indexing, search structures, and query optimization | LLM Application Foundations |
| 3 | Agentic Systems & Harness Engineering | Model Context Protocol (MCP) & Context Engineering | Computer Networks — protocols, client-server architecture, and RPC | Dynamic Inline Tooling, MCP & Cache-Preserving Toolmaking |
| 3 | AI Infrastructure & Inference | FlashAttention & Subquadratic Sparse Attention | Operating Systems — I/O scheduling and memory-bound versus compute-bound work | GPU Architecture & the Memory Wall |
| 3 | Applied ML & Model Engineering | Transformers, Attention & Pretraining | Machine Learning — sequence modeling and representation learning | Deep Learning Foundations |
| 3 | Frontier Systems | Consensus & Coordination | Distributed Systems — consensus, replication, and fault tolerance | Distributed Systems Foundations |
| 3 | Multimodal AI & Embodied World Modeling | Diffusion, Flow-Matching & Generative Video Foundations | Probability & Statistics — stochastic processes and generative modeling | None |
| 3 | Production AI Products | Dynamic Context Engineering, KV-Cache Diagnostics & Prefix Breakpoints | Databases — API/schema versioning, semantic routing, and transactional isolation | Embeddings & Vector Search |
| 4 | Agentic Systems & Harness Engineering | Planning, Adaptive Reasoning Scaling & Test-Time Compute | Artificial Intelligence — search, planning, and state-space reasoning | Dynamic Inline Tooling, MCP & Cache-Preserving Toolmaking |
| 4 | AI Infrastructure & Inference | KV-Cache Management, RadixAttention & Multi-Head Latent Attention | Operating Systems — virtual memory, paging, and fragmentation | FlashAttention & Subquadratic Sparse Attention |
| 4 | AI Safety, Alignment & Interpretability | Constitutional AI & Scalable Oversight | Distributed Systems — delegation, trust, and verification of work you cannot fully check | RLHF, Direct Alignment & Verifiable Reasoning (GRPO) |
| 4 | Applied ML & Model Engineering | Pre-Training Data, Synthetic Mid-Training & Scaling Laws | Databases — data cleaning, deduplication, and ETL pipelines | Transformers, Attention & Pretraining |
| 4 | Frontier Systems | Distributed Training & Parallelism | Parallel Computing — collective communication and parallel decomposition | Consensus & Coordination |
| 4 | Land the Elite AI Role | Building a Portfolio That Gets Noticed | , | Running Experiments, Ablations & Tracking |
| 4 | Production AI Products | Production RAG: Asymmetric Late Interaction & Contextual Retrieval | Databases — query processing, joins, and information retrieval | Embeddings & Vector Search |
| 5 | Agentic Systems & Harness Engineering | Context Hubs, JIT Context Delivery & Semantic Routers | Operating Systems — memory hierarchy, paging, and caching | Planning, Adaptive Reasoning Scaling & Test-Time Compute |
| 5 | AI Infrastructure & Inference | Continuous Batching & Throughput Scheduling | Operating Systems — CPU scheduling, throughput versus latency tradeoffs | KV-Cache Management, RadixAttention & Multi-Head Latent Attention |
| 5 | AI Safety, Alignment & Interpretability | Mechanistic Interpretability Foundations | Compilers — reverse engineering, intermediate representations, and program analysis | The Alignment Problem & Safety Foundations |
| 5 | Applied ML & Model Engineering | Hybrid Architectures: SSM-Transformer Hybrids & Attention Alternatives | Computer Architecture — specialized processors and hardware-agnostic compilation | Pre-Training Data, Synthetic Mid-Training & Scaling Laws |
| 5 | Multimodal AI & Embodied World Modeling | Latent Diffusion & Conditioning | Computer Vision — image synthesis and conditional generation | Diffusion, Flow-Matching & Generative Video Foundations, Vision-Language Model Architecture |
| 5 | Production AI Products | Evaluation-Driven Development (EDD) & Synthetic Data Harnesses | Software Engineering — automated testing and continuous integration | Production RAG: Asymmetric Late Interaction & Contextual Retrieval |
| 6 | Agentic Systems & Harness Engineering | Hardware-Isolated Sandboxing, Out-of-Band Watchdogs & Kernel Confinement | Operating Systems — virtualization, namespaces, and process isolation | Context Hubs, JIT Context Delivery & Semantic Routers |
| 6 | Applied ML & Model Engineering | Supervised Fine-Tuning | Machine Learning — transfer learning and supervised training | Pre-Training Data, Synthetic Mid-Training & Scaling Laws |
| 6 | Frontier Systems | GPU Cluster Scheduling | Operating Systems — scheduling, resource allocation, and fairness | Distributed Training & Parallelism |
| 6 | Land the Elite AI Role | Technical Writing & Communication | , | Building a Portfolio That Gets Noticed |
| 6 | Multimodal AI & Embodied World Modeling | Video & Temporal Generative Models | Signal Processing — temporal signals, sampling, and reconstruction | Latent Diffusion & Conditioning |
| 7 | AI Infrastructure & Inference | Quantization & Model Compression | Computer Architecture — number representation and fixed-point arithmetic | Continuous Batching & Throughput Scheduling |
| 7 | AI Safety, Alignment & Interpretability | Interpretability in Practice: SAEs, NLAs & Activation Patching | Software Engineering — instrumentation, debugging, and observability of complex systems | Mechanistic Interpretability Foundations |
| 7 | Applied ML & Model Engineering | Parameter-Efficient Fine-Tuning: LoRA & QLoRA | Linear Algebra — matrix rank, decomposition, and low-rank approximation | Supervised Fine-Tuning |
| 7 | Frontier Systems | Real-Time & Streaming AI Systems | Distributed Systems — event-driven architecture and stream processing | GPU Cluster Scheduling |
| 7 | Production AI Products | AI Observability & Tracing | Software Engineering — monitoring, logging, and distributed tracing | Evaluation-Driven Development (EDD) & Synthetic Data Harnesses |
| 8 | Agentic Systems & Harness Engineering | Self-Organizing Agent Teams, Dynamic Role Negotiation & Step Auditing | Distributed Systems — coordination, message passing, and consensus | Hardware-Isolated Sandboxing, Out-of-Band Watchdogs & Kernel Confinement |
| 8 | AI Infrastructure & Inference | Speculative Decoding & Latency Optimization | Computer Architecture — speculative and out-of-order execution | Quantization & Model Compression |
| 8 | AI Safety, Alignment & Interpretability | Safety Evaluations, Containment Auditing & Agent Red-Teaming | Software Engineering — test design, coverage, and validation under uncertainty | Constitutional AI & Scalable Oversight |
| 8 | Land the Elite AI Role | Open-Source Contribution & Visibility | , | Technical Writing & Communication |
| 8 | Multimodal AI & Embodied World Modeling | Spatial Intelligence, World Action Models & Physical Simulation | Machine Learning — sequence modeling and unified representations | Video & Temporal Generative Models |
| 9 | Agentic Systems & Harness Engineering | Software 3.0: Intent Specification, Harness Engineering & Autonomous Verification Loops | Compilers — program optimization, intermediate representations, and iterative refinement loops | Planning, Adaptive Reasoning Scaling & Test-Time Compute |
| 9 | AI Safety, Alignment & Interpretability | Dual-Use Cyber Capabilities, Exploit Chaining & Red-Teaming Depth | Information Security — adversarial thinking, threat modeling, and penetration testing | Safety Evaluations, Containment Auditing & Agent Red-Teaming |
| 9 | Applied ML & Model Engineering | Preference Optimization, RLVR & Machine-Checked Autoformalization | Machine Learning — reinforcement learning and policy optimization | Parameter-Efficient Fine-Tuning: LoRA & QLoRA |
| 9 | Frontier Systems | Vector Databases at Scale | Databases — indexing, sharding, and query optimization | Real-Time & Streaming AI Systems |
| 9 | Land the Elite AI Role | The Elite AI Hiring Pipeline | , | Open-Source Contribution & Visibility |
| 9 | Multimodal AI & Embodied World Modeling | Multimodal Training & Instruction Tuning | Machine Learning — transfer learning and supervised fine-tuning | Spatial Intelligence, World Action Models & Physical Simulation |
| 9 | Production AI Products | AI Security & Red-Teaming | Information Security — threat modeling, penetration testing, and OWASP | AI Observability & Tracing |
| 10 | Agentic Systems & Harness Engineering | Agent Verification Loops & HITL Guardrails | Software Engineering — testing, regression suites, and observability | Self-Organizing Agent Teams, Dynamic Role Negotiation & Step Auditing |
| 10 | AI Infrastructure & Inference | Production Serving, PD Disaggregation & Autoscaling | Distributed Systems — load balancing, replication, and capacity planning | Speculative Decoding & Latency Optimization |
| 10 | AI Safety, Alignment & Interpretability | AI Regulatory Engineering: Statutory Auditing, EU AI Act & Non-Delegation Controls | Information Security — policy, compliance, and audit-ready documentation | Safety Evaluations, Containment Auditing & Agent Red-Teaming |
| 10 | Applied ML & Model Engineering | Knowledge Distillation & Model Compression | Machine Learning — model compression and the teacher-student paradigm | Preference Optimization, RLVR & Machine-Checked Autoformalization |
| 10 | Production AI Products | LLMOps, Cost Governance & Geopolitical Fallbacks | Software Engineering — release management, versioning, and CI/CD | AI Security & Red-Teaming |
| 11 | Agentic Systems & Harness Engineering | Agent Runtime Security: Threat Modeling, Tool Hijacking & Credential Isolation | Information Security — threat modeling, capability isolation, and least privilege | Agent Verification Loops & HITL Guardrails |
| 11 | AI Infrastructure & Inference | Sovereign AI, MLX & Local GPU Clustering | Computer Architecture — specialized processors and hardware-agnostic compilation | Production Serving, PD Disaggregation & Autoscaling |
| 11 | AI Safety, Alignment & Interpretability | Research Methodology: Reading & Reproducing Papers | Software Engineering — the scientific method, controlled experiments, and reproducibility | Interpretability in Practice: SAEs, NLAs & Activation Patching |
| 11 | Applied ML & Model Engineering | Rigorous Model Evaluation | Statistics — sampling, confidence intervals, and experimental design | Knowledge Distillation & Model Compression |
| 11 | Frontier Systems | Fault Tolerance & Resilient Operations | Distributed Systems — fault tolerance, checkpointing, and recovery | Vector Databases at Scale |
| 11 | Land the Elite AI Role | Compensation & Negotiation | , | The Elite AI Hiring Pipeline |
| 11 | Multimodal AI & Embodied World Modeling | Multimodal Inference & Evaluation | Machine Learning — model evaluation and benchmark design | Multimodal Training & Instruction Tuning |
| 12 | Frontier Systems | Cluster Observability & Capacity Planning | Distributed Systems — monitoring, capacity planning, and performance analysis | Fault Tolerance & Resilient Operations |
| 12 | Production AI Products | Shipping & Operating AI Products | Software Engineering — deployment, resilience, and product engineering | LLMOps, Cost Governance & Geopolitical Fallbacks |