Phase 19
Capstone Projects
Prove everything you learned. Build portfolio-grade systems.
Lessons (85)
- 01Capstone 01 — Terminal-Native Coding Agent
- 02Capstone 02 — RAG over Codebase (Cross-Repo Semantic Search)
- 03Capstone 03 — Real-Time Voice Assistant (ASR to LLM to TTS)
- 04Capstone 04 — Multimodal Document QA (Vision-First PDF, Tables, Charts)
- 05Capstone 05 — Autonomous Research Agent (AI-Scientist Class)
- 06Capstone 06 — DevOps Troubleshooting Agent for Kubernetes
- 07Capstone 07 — End-to-End Fine-Tuning Pipeline (Data to SFT to DPO to Serve)
- 08Capstone 08 — Production RAG Chatbot for a Regulated Vertical
- 09Capstone 09 — Code Migration Agent (Repo-Level Language / Runtime Upgrade)
- 10Capstone 10 — Multi-Agent Software Engineering Team
- 11Capstone 11 — LLM Observability & Eval Dashboard
- 12Capstone 12 — Video Understanding Pipeline (Scene, QA, Search)
- 13Capstone 13: Stateless MCP Server with Registry and Governance
- 14Capstone 14 — Speculative-Decoding Inference Server
- 15Capstone 15 — Constitutional Safety Harness + Red-Team Range
- 16Capstone 16 — GitHub Issue-to-PR Autonomous Agent
- 17Capstone 17 — Personal AI Tutor (Adaptive, Multimodal, with Memory)
- 20Agent Harness Loop Contract
- 21Tool Registry with Schema Validation
- 22JSON-RPC 2.0 Over Newline-Delimited Stdio
- 23Function Call Dispatcher
- 24Plan-Execute Control Flow
- 25Capstone Lesson 25: Verification Gates and the Observation Budget
- 26Capstone Lesson 26: Sandbox Runner with Denylist and Path Jail
- 27Capstone Lesson 27: Eval Harness with Fixture Tasks
- 28Capstone Lesson 28: Observability with OTel GenAI Spans and Prometheus Metrics
- 29Capstone Lesson 29: End-to-End Coding Agent on the Harness
- 30BPE Tokenizer From Scratch
- 31Tokenized Dataset with Sliding Window
- 32Token and Positional Embeddings
- 33Multi-Head Self-Attention
- 34Transformer Block from Scratch
- 35GPT Model Assembly
- 36Training Loop and Evaluation
- 37Loading Pretrained Weights
- 38Capstone Lesson 38: Classifier Fine-Tuning by Head Swap
- 39Capstone Lesson 39: Instruction Tuning by Supervised Fine-Tuning
- 40Capstone Lesson 40: Direct Preference Optimization from Scratch
- 41Capstone Lesson 41: Full Evaluation Pipeline
- 42Large Corpus Downloader
- 43HDF5 Tokenized Corpus
- 44Cosine LR with Linear Warmup
- 45Gradient Clipping and Mixed Precision
- 46Gradient Accumulation
- 47Checkpoint Save and Resume
- 48Distributed Data Parallel and FSDP from Scratch
- 49Language Model Evaluation Harness
- 50Hypothesis Generator
- 51Literature Retrieval
- 52Experiment Runner
- 53Result Evaluator
- 54Paper Writer
- 55Critic Loop
- 56Iteration Scheduler
- 57End-to-End Research Demo
- 58Vision Encoder Patches
- 59Vision Transformer Encoder
- 60Projection Layer for Modality Alignment
- 61Cross-Attention Fusion
- 62Vision-Language Pretraining
- 63Multimodal Evaluation
- 64Chunking Strategies, Compared
- 65Hybrid Retrieval with BM25 and Dense Embeddings
- 66Cross-Encoder Reranker
- 67Query Rewriting: HyDE, Multi-Query, and Decomposition
- 68RAG Evaluation: Precision, Recall, MRR, nDCG, Faithfulness, Answer Relevance
- 69End-to-End RAG System
- 70Task Spec Format
- 71Classical Metrics
- 72Code Exec Metric
- 73Perplexity and Calibration
- 74Leaderboard Aggregation
- 75End-to-End Eval Runner
- 76Collective Ops From Scratch
- 77Data Parallel DDP From Scratch
- 78ZeRO Optimizer State Sharding
- 79Pipeline Parallel and Bubble Analysis
- 80Sharded Checkpoint and Atomic Resume
- 81End-to-End Distributed Training
- 82Capstone 82 — Jailbreak Taxonomy
- 83Capstone 83 — Prompt Injection Detector
- 84Capstone 84 — Refusal Evaluation
- 85Capstone 85 — Content Classifier Integration
- 86Capstone 86 — Constitutional Rules Engine
- 87Capstone 87 — End-to-End Safety Gate