Knowledge Base
A growing set of architecture notes and production playbooks. Each route is designed as a reusable reference for system design, implementation decisions, and engineering trade-offs.
Agentic AI System Design
Complete production reference: single vs multi-agent systems, model routing, tool contracts, memory, orchestration, evaluation, approval gates, reliability, cost, RAG, observability, security, and privacy.
[02]AI Evals
A practical guide to evaluating AI agents: what evals measure, how they differ from benchmarks, which metrics to track per task, and the evaluation loop that separates teams shipping production AI.
[03]LLM Fine-Tuning
The complete guide to fine-tuning: when it pays off, dataset preparation, LoRA and QLoRA, hyperparameters, and preference alignment with RLHF or DPO.
[04]MCP Complete Explanation
The Model Context Protocol in full: the drawer-of-chargers problem it solves, host, client, and server, where MCP sits inside an agent, local vs remote servers, security, and building your first server.
[05]RAG Explained
Retrieval-augmented generation in full: how RAG works inside the architecture, the misconceptions that persist, ten RAG patterns for 2026, and where to go next.
[06]Harness Engineering
The discipline behind production AI: why the model is the easy part, and how testing, tracing, guardrails, and continuous evaluation build agentic systems people can rely on.
[07]Vector Databases
Embeddings, indexing (HNSW, IVF), hybrid search, and RAG — the complete 2026 guide to the unglamorous infrastructure behind chatbots with memory and semantic search.
[08]Nvidia CUDA in 100 Seconds
The parallel computing platform behind GPU training: why GPUs win at parallel work, the host-device execution model, and a first CUDA program in C++.
[09]React Architecture
A practical decision guide for component boundaries, state ownership, rendering, data fetching, performance, accessibility, and maintainable frontend systems.
[10]Microfrontend Architecture
How to split independently delivered frontend domains while keeping contracts, routing, shared dependencies, design systems, and observability under control.
[11]Enterprise Design Systems
Tokens, component contracts, Web Components, Storybook, accessibility, framework parity, governance, and release strategies for large design-system platforms.