Long-form AI engineering articles for systems, workflows, architecture, and production decisions.
Choose a topic to focus the library.
48 articles found
JSON, XML, YAML, TOML, and TOON each carry different signals inside a prompt. Knowing what the model actually does with each format is how you make format a deliberate engineering decision instead of an afterthought.
Use structured JSON inputs and outputs to make LLM systems predictable with typed parsing, validation, and deterministic routing.
The shift from classical ML to modern AI is primarily architectural. Discover why engineering is moving from static algorithms to orchestrated systems.
Mastering prompt engineering as a strategic discipline for guiding LLMs toward precise, reliable outputs in enterprise workflows.
A complete 2026 roadmap for mastering AI Engineering covering fundamentals, LLMs, RAG, agents, evaluation, and production deployment.
Treat RAG embedding indexes as deployable artifacts with current and candidate manifests, recall regression checks, metadata validation, snapshots, restore drills, lineage, and promotion evidence.
Build a local Feast feature store with point-in-time training retrieval, SQLite online serving, freshness gates, and deterministic scoring for hybrid AI decisions.
Build a .NET data-contract pipeline that admits only fresh, owned, audience-safe, lineage-backed insurance policy documents into RAG or agent context.
Treat long-running agent context as a controlled system. Pin required facts, compact old history, expire stale notes, and carry forward only what the next turn is allowed to rely on.
Build a local-first memory architecture with deterministic write guardrails, bounded retrieval, context budgeting, and evidence-only answers.
Build a deterministic context budgeting pipeline that prioritizes required evidence, enforces token limits, and composes inspectable prompts.
Learn the core AI pattern for semantic retrieval with embeddings as representation, deterministic vector ranking, thresholded control, and optional pgvector scaling.
Build a local semantic search engine for engineering runbooks using Ollama and Microsoft.Extensions.AI for deterministic retrieval.
Designing, structuring, and managing context to deliver personalized, consistent, memory-aware AI across sessions and complex workflows.
Build a local, category-aware RAG system with ASP.NET Core MVC, Ollama LLMs, and PostgreSQL + pgvector for secure, context-bound enterprise Q&A.
Build a C# console-based Retrieval-Augmented Generation (RAG) system leveraging Semantic Kernel, Ollama LLMs, and QDrant vector search for context-aware document Q&A.
Build a Python expense-policy workflow that uses Jev for typed semantic decisions while deterministic code owns amounts, dates, duplicates, thresholds, audit records, and final routing.
Loop engineering moves AI work from one prompt at a time to bounded control systems with triggers, gates, state, budgets, and escalation paths.
A gateway that discovers, validates, and delegates to a remote specialist over A2A. Both agents run on Microsoft Foundry and deploy to Azure Container Apps from one image.
Build a durable local-first agent workflow in C# with narrow roles, deterministic policy-bounded execution, live runtime logs, and restart-safe state, using home repair as the concrete example.
Build a grounded local-first multi-agent research workflow in C# with deterministic retrieval, narrow agent roles, structured JSON contracts, and final citation validation.
Build a local-first incident triage agent with Microsoft Agent Framework using structured workflows, deterministic policy enforcement, and explicit runtime guardrails.
Build a local-first .NET intent router that keeps routing, tools, and safety deterministic while using embeddings only for disambiguation under controlled thresholds.
Build a reactive local AI agent using Semantic Kernel and Ollama for deterministic tool execution and natural language reasoning.
Build a production-ready MCP setup in C# and .NET 9, integrating a simple time tool with Ollama for deterministic AI workflows.
Dive into Semantic Kernel for orchestrating AI and deterministic code in hybrid workflows, with a deep dive into a sales analytics application.
Build a local Python runtime-control layer that uses feature flags to select model routes, prompt versions, tool authority, retrieval corpora, reasoning modes, and kill switches before an AI workflow executes.
Turn chain-of-thought from a prompt habit into an AI engineering contract with private reasoning modes, visible rationales, evidence, logging policy, and evaluation checks.
Build a Python tool-result admission gate that rejects unsafe tool output, redacts sensitive fields, packs only admitted context, and uses Microsoft Foundry for final synthesis.
Build a markdown-first governance package that treats golden eval cases as controlled production assets with labeling rules, adjudication, versioning, leakage controls, and release evidence.
Build a Python moderation gate that scores text with Detoxify, asks a local Ollama model for structured review when policy requires it, and keeps the final content-risk decision in deterministic code.
Define a production prompt and context lineage contract that records effective model inputs, protects sensitive content, and supports evidence-based incident replay.
Build a local C# approval workflow where risky AI-proposed actions create explicit approval requests, consume scoped expiring tokens, and leave behind auditable decision records.
Treat live MCP servers as versioned tool contracts. Discover live schemas, replay frozen probes, and block risky drift before promotion.
Treat model swaps as controlled experiments: frozen evals, shadow replay, deterministic scoring, and explicit promotion gates before an AI system changes models.
Build a support agent workflow in C# with trusted vs untrusted content labels, deterministic tool authorization, approval-gated sensitive actions, and redacted audit storage.
Build a local-first prompt experiment harness with frozen eval cases, deterministic scoring, live LM Studio execution, and evidence-based prompt promotion.
Build an evaluation harness with telemetry and feedback loops to gate local AI releases with measurable reliability. Track quality over time, catch regressions early, and ship with clear deployment thresholds.
Build a minimal but production-aligned guardrailed AI assistant in C# that runs fully locally using Ollama, focusing on clear boundaries, strict contracts, deterministic execution, and safe failure modes.
Build a .NET batch inference runner with request manifests, idempotency keys, retry envelopes, checkpoints, failed-row quarantine, and replayable output records backed by Microsoft Foundry.
Build a .NET streaming AI gateway with server-sent events, cancellation handling, partial-output policy, and finalization records before streamed text is treated as committed.
Build a local-first inference gateway with exact cache keys, per-tenant budget reservation, sync-vs-batch routing, deferred execution, and measurable cost controls in C#.
Build a local-first inference gateway with deterministic rate limiting, queue backpressure, retries, timeouts, circuit breakers, fallback routing, and measurable runtime reliability metrics.
A practical guide to building an AI engineering portfolio that demonstrates system judgment, operational tradeoffs, and production readiness without relying on flashy demos.
Design a markdown-first documentation context package with llms.txt, source ownership, freshness rules, exclusions, and eval questions so agents can use docs without treating the whole site as context.
Repository instruction files are becoming the control plane for AI-assisted coding: a short AGENTS.md map, deep docs as the system of record, and task-specific specs for execution.
Spec-driven AI-assisted coding treats specifications as execution contracts, constraining agent behavior for predictable, production-grade outcomes.
Explore emerging trends in software engineering, AI-assisted tools, and a step-by-step guide to setting up a private AI coding environment.
How the library works
The library collects long-form technical writing from The Modern Engineer, organized around AI engineering, system design, architecture, workflows, and production decisions.
The newsletter archive is chronological. The article library is organized for browsing by topic, category, search, and read status.
New writing is added as newsletter issues are published and categorized for the library.
Yes. Read status is stored locally in this browser, and you can export or import your read-history data from Manage Read History.