The FDE concept map
167 concepts across 10 tracks, drawn with the 545 links between them. 185 of those links (34%) cross tracks, which is the part worth looking at: the concepts that decide FDE interviews are rarely the ones that sit neatly inside one topic. Hover a concept to see only what it touches. Click to read it.
same trackcrosses tracks free membersbigger dot = more connections167 shown
Where the map is dense
The most-connected concepts are the ones the rest of the library keeps reaching for, which makes them the highest-leverage things to be solid on: Retrieval-Augmented Generation (RAG) (24), Golden Datasets and Eval Sets (18), Latency Optimization (16), Guardrails (15), and Observability for AI Systems (15). If you are deciding where to spend a week, start with a hub rather than a leaf.
Every concept, by track
Foundations of LLMs & GenAI30
- Embeddings & Vector Representations13
- KV Cache13
- The Transformer, Intuitively9
- Model Routing and Cascades9
- The Context Window8
- Why LLMs Hallucinate8
- RLHF (Alignment)8
- Reward Models8
- Prompt Engineering7
- Fine-tuning vs RAG vs Prompting7
- Structured Output and Schema Validation7
- Tokenization & Tokens6
- Temperature, Top-p and Sampling6
- Multimodal Models and VLMs6
- Direct Preference Optimization (DPO)5
- Inference-Time Compute5
- Prompt Caching and Semantic Caching5
- Model Selection for Enterprise Deployments5
- Attention and Self-Attention4
- Constrained Decoding4
- Constitutional AI and RLAIF4
- Policy Optimization: PPO and GRPO4
- RoPE and Positional Encodings3
- Chain-of-Thought Prompting3
- LoRA and Parameter-Efficient Fine-tuning3
- Mixture of Experts (MoE)3
- Scaling Laws3
- Speech and Voice AI3
- Autoregressive Decoding3
- Diffusion Models2
Retrieval & Agents31
- Retrieval-Augmented Generation (RAG)24
- Guardrails15
- Tool / Function Calling12
- Vector Databases11
- MCP (Model Context Protocol)11
- The ReAct Loop (Reason, Act, Observe)11
- Chunking Strategies10
- AI Agents and Tool Use10
- Agentic Evals: Grading the Trajectory, Not the Answer9
- Bounded Autonomy and Human-in-the-Loop9
- Hybrid Search (Lexical + Vector)8
- Reranking and Two-Stage Retrieval8
- Agent vs Workflow vs a Single Call8
- Permission-Aware RAG8
- Agent Memory7
- Agent-to-Agent Interoperability (A2A)7
- Approximate Nearest Neighbor (ANN)6
- Agent Frameworks and How Agents Fail6
- Embedding Versions and Drift6
- Multi-Agent Orchestration5
- Context Window Management for FDE Agents5
- Context Failure Modes5
- GraphRAG and Contextual Retrieval5
- Query Rewriting, Expansion and HyDE5
- Index Freshness and Staleness Windows5
- Late-Interaction Retrieval (ColBERT)5
- AG-UI: The Agent-User Interaction Protocol5
- AP2: The Agent Payments Protocol5
- Text-to-SQL5
- Document Parsing and Extraction5
- TF-IDF and BM253
Evaluation & ML Foundations24
- Golden Datasets and Eval Sets18
- LLM-as-a-Judge11
- Precision, Recall and F110
- Offline vs Online Evaluation10
- A/B, Canary and Shadow Testing9
- Gradient Descent & Learning Rate8
- Bias-Variance Tradeoff7
- Information Theory for ML: Entropy, Cross-Entropy, KL and Perplexity6
- Overfitting and Regularization6
- Evaluating RAG Systems6
- Calibration and Uncertainty5
- Benchmarks and Their Limits5
- Faithfulness vs Answer Relevancy5
- Synthetic Data Generation4
- Catastrophic Forgetting3
- Loss Functions3
- Activation Functions3
- Neural Network Basics: Perceptron to MLP3
- Semi-Supervised and Self-Training3
- Computer Vision: Classification, Detection, Segmentation3
- Multi-Armed Bandits2
- Normalization: Batch vs Layer2
- Handling Imbalanced Data2
- Convex vs Non-Convex Optimization2
System Design for AI in Production20
- Latency Optimization16
- Observability for AI Systems15
- VPC and Air-Gapped Deployment13
- From Proof-of-Concept to Production11
- Circuit Breakers and Backpressure10
- The Ontology (Semantic Layer)10
- Idempotency8
- AI Cost and Unit Economics7
- Retries, Exponential Backoff and Jitter7
- Rate Limiting7
- The Walking Skeleton (Thin Slice First)6
- REST API Design for Integrations6
- Containers and Kubernetes for Customer Deployments6
- Consistency, CAP and What Your Workflow Actually Needs6
- Deployment Models: SaaS, BYOC, On-Prem and Air-Gapped6
- Message Queues and Pub/Sub5
- Fallbacks and Provider Failover5
- Palantir's Platform: Foundry, AIP, Gotham and Apollo5
- The Enterprise AI Reference Architecture5
- SLOs, SLIs and Error Budgets5
Data & SQL Engineering14
- Data Quality and Validation14
- Idempotent Data Pipelines12
- ETL vs ELT10
- Query Optimization and Execution Plans10
- Change Data Capture (CDC)8
- Lakehouse, Delta and the Medallion Architecture8
- Batch vs Streaming7
- Dimensional Modeling and Slowly Changing Dimensions6
- Orchestration: DAGs, Retries and Backfills6
- Spark Internals and Performance Tuning5
- SQL vs NoSQL: Choosing a Data Store5
- Deduplication and LSH4
- SQL Window Functions3
- Gaps and Islands3
AI Security, Privacy & Governance16
- RBAC, ABAC and Identity at Runtime for Agents13
- Multi-Tenancy and Data Isolation12
- Audit Trails and Traceability11
- OAuth 2.0, JWTs and Service Credentials9
- PII Handling and Redaction8
- IAM and Least Privilege8
- AI Incident Response8
- Prompt Injection and Defense7
- AI Governance (SOC2, EU AI Act)7
- Enterprise SSO: SAML and OIDC7
- Data Residency and Sovereignty7
- Secrets Management7
- Agent Sandboxing and Execution Isolation5
- Differential Privacy3
- Federated Learning3
- Mechanistic Interpretability3
The map is for orientation. If you would rather be told what to do in order, the start-here path sequences the same material by background and stage, and the courses walk it front to back.