A living library for engineers building with AI
Featured
Production RAG with LangGraph and Langfuse: Part 1, Every Stage as a Traceable Graph
LLM Fine-Tuning End to End | Part 1: Methods, Data, and Training Optimization
Production AI System Design for Millions of Users: Part 1, The Architecture
AI & Machine Learning
Models, inference, MLOps, applied AI, and learning notes from the field.
What Mistral AI Provides for Developers: The Full Ecosystem, Models, and APIs
A complete developer's map of the Mistral AI ecosystem in 2026: the model lineup, La Plateforme API, the Agents API and connectors, three fine-tuning routes, embeddings and document AI, and how an AIML engineer actually builds with each.
What Qwen Provides for Developers: Alibaba's Open Model Ecosystem, APIs, and Agents
A complete developer's map of Alibaba's Qwen ecosystem in 2026: the sprawling model lineup, the DashScope and Model Studio API, the Qwen-Agent framework, fine-tuning routes, embeddings and multimodal building blocks, and how an AIML engineer actually builds with each.
MLOps Basics for Small Teams
A lightweight view of versioning data, tracking experiments, monitoring models, and shipping machine learning without heavy platform overhead.
Fine-Tuning vs RAG for AI Products
A practical decision guide on when to retrieve context, when to adapt a model through training, and why most teams should start with evaluation before choosing either.
Engineering
Systems, infrastructure, databases, indexing, and product architecture.
Observability for AI Systems
An engineering guide to tracing prompts, retrieval steps, model latency, token usage, errors, and quality signals in AI-backed products.
API Rate Limiting: System Design Notes
A systems-focused breakdown of token buckets, fixed windows, sliding windows, distributed counters, and abuse protection for production APIs.
Designing Reliable Job Queues
Engineering notes on retries, idempotency, dead-letter queues, backpressure, and observability for background work that cannot afford to silently fail.
System Design for AI Products: What Changes After LLMs
A practical look at what breaks in traditional product architecture when LLMs enter the stack, and how to design for slow calls, uncertain outputs, retrieval quality, and new failure modes.
How Things Work
Deep explainers on internals, architecture, platforms, and technical ideas.
How CDNs Make Web Apps Feel Fast
A practical explanation of how CDNs use edge caching, cache keys, headers, and revalidation to make modern web apps load faster without losing freshness.
How GPUs Run AI Workloads
A plain-language breakdown of parallel computation, tensor operations, memory bandwidth, and why GPUs became central to modern AI training and inference.
How Indexing Works in Databases
A practical mental model for B-trees, hash indexes, query planners, and why indexes speed reads while adding write and storage cost.
How Vector Search Works: From Embeddings to Retrieval
A practical explanation of embeddings, similarity metrics, nearest-neighbor indexes, and retrieval tradeoffs for anyone building AI search systems.
Guides
Practical walkthroughs, implementation notes, and build-focused tutorials.
Production AI Agents with LangGraph and Langfuse: Part 2, Tools, Human-in-the-Loop, and Traceable Loops
A phase-by-phase guide to building production-grade AI agents with LangGraph: state, routing, reasoning, tool execution, the agent loop, human-in-the-loop interrupts, memory, and multi-agent subgraphs, each step traced in Langfuse.
LLM Fine-Tuning End to End | Part 1: Methods, Data, and Training Optimization
A comprehensive technical guide to fine-tuning large language models: what it is, when to use it, the full method spectrum from continued pretraining to DPO, PEFT techniques like LoRA and QLoRA, dataset construction, and the optimization strategies that make training fit on real hardware.
LLM Fine-Tuning, End to End | Part 2: Serving a Fine-Tuned Model to Billions Efficiently
A comprehensive technical guide to deploying fine-tuned LLMs at scale: inference fundamentals, continuous batching, PagedAttention, quantization, speculative decoding, multi-LoRA serving, autoscaling, cost control, and production monitoring.
Production AI System Design for Millions of Users: Part 1, The Architecture
A complete architecture for AI products at scale: the traditional system-design layers that still apply (caching, queues, autoscaling, CI/CD, reliability) plus the new AI-specific control plane (LLM gateway, semantic cache, retrieval, guardrails, evals, and observability).
Comparisons
Tradeoffs, tool choices, framework comparisons, and decision guides.
Next.js vs React Router v7 (Remix): Choosing for Product Apps
A practical comparison of Next.js and React Router v7 (formerly Remix) across routing, data loading, forms, deployment, and team ergonomics for product teams.
Redis Streams vs Kafka: Choosing for Event Systems
A practical comparison of Redis Streams and Apache Kafka across throughput, retention, replay, consumer groups, operational cost, and product complexity for engineering teams.
LangChain vs LlamaIndex for RAG: Which Framework Should You Use?
A practical comparison of LangChain and LlamaIndex for building RAG systems, covering orchestration, retrieval, observability, integrations, and real-world fit.
Pinecone vs Weaviate vs Qdrant: Choosing a Vector Database
A practical comparison of Pinecone, Weaviate, and Qdrant across retrieval quality, filtering, hybrid search, pricing, and operational fit for production RAG systems.
Research & Experiments
Benchmarks, evaluations, prototypes, and exploratory technical work.
Experimenting With Hybrid Search
A structured experiment comparing keyword, vector, and hybrid retrieval across exact-term, conceptual, and mixed queries, with methodology, expected patterns, and practical guidance.
Evaluating Small Agent Workflows: When They Help and When They Don't
An experiment on when lightweight AI agents help developer workflows, when they fail, and which signals are worth measuring, including a reproducible methodology you can adapt.
Benchmarking Vector Database Retrieval
A reproducible methodology for comparing recall, latency, metadata filtering, and developer ergonomics across vector search systems, with honest limitations and practical guidance.
Evaluating RAG Quality: A Small Retrieval Experiment
An experiment-driven look at how chunk size, metadata filtering, and reranking affect RAG retrieval quality, and why measurement is non-negotiable.
