AI Engineering
Cut Your AI Coding Bill With Model Routing
Most agent tokens aren't thinking — they're I/O. Here's a free local setup that routes grunt work to cheap models.
September 13, 2026 · 12 min read · AI Agents · Agentic Systems
By Vinod Kumar · Staff Engineer at ServiceNow, building production AI systems
AI engineering is a craft. These are the notes from the build floor.
Deep technical writing on agents, context engineering, RAG, and inference — written by someone who ships these systems, not just writes about them. No hype. No tutorials. Just honest architecture decisions from production.
13 in-depth notes · 8 topics · 3 guided series
Featured
AI Agents
AI Agents Are Becoming Software Systems
Why building reliable agents requires systems engineering, not just prompting.
12 min read · September 12, 2026
Read articleLatest engineering notes
All articlesAI Agents
Building a Private Local AI Agent with Ollama
A 400-line agent with no framework: local inference, jailed tools, and the four real bugs that evals caught.
September 13, 2026 · 15 min read · AI Agents · Ollama
Knowledge Graphs
What Is a Knowledge Graph?
Entities, relations, and why structured knowledge grounds AI systems that vectors alone can't.
September 12, 2026 · 9 min read · Knowledge Graphs · RAG
Agentic Systems
Harness Engineering: The Discipline Behind Reliable Agents
The model proposes, the harness disposes — budgets, tools, evals, and the loop that makes agents production-worthy.
September 12, 2026 · 11 min read · Agentic Systems · AI Agents
Context Engineering
Context Engines: The System That Feeds the Model
From scattered sources to one assembled, budgeted, trustworthy context — how to build the engine.
September 12, 2026 · 10 min read · Context Engineering · RAG
Explore the AI engineering stack
All topicsAI Engineering
- AI Engineering9 notesPatterns and practices for building AI-powered software that actually works.
- AI Agents6 notesDesigning agents as software systems — planning, tools, memory, and control loops.
- Agentic Systems4 notesMulti-agent coordination, workflows, evaluation, and reliability.
- Context Engineering2 notesThe missing layer: assembling the right context for every inference.
- RAG5 notesRetrieval-augmented generation — chunking, retrieval, reranking, and grounding.
- LLM Inference1 noteHow LLM serving works — batching, KV-cache, quantization, and latency.
- Knowledge Graphs3 notesStructured knowledge, entities, relations, and graph-grounded reasoning.
- AI Infrastructure1 noteModel serving, GPUs, scaling, and the platform behind production AI.
Start here
Guided seriesA practical path into AI Engineering.
New to the field? Read in order. Each step links to the best place to begin — then branch into the stack above.
Step 01 · AI Engineering
AI Engineering Foundations
How modern AI systems fit together — and which of your engineering instincts still apply.
Start withWhat Software Engineers Need to Know About AI Engineering8 minStep 02 · Context Engineering
LLMs & Context Engineering
How models consume information — and how to assemble the right context for every inference.
Start withContext Engineering: The Missing Layer in AI Applications10 minStep 03 · RAG
RAG & Knowledge
Retrieval, chunking, ranking, embeddings — and the structured knowledge underneath.
Start withRAG Is a Retrieval Problem Before It Is an LLM Problem9 minStep 04 · AI Agents
AI Agents
Tools, memory, planning, and evaluation — agents as software systems, not prompts.
Start withAI Agents Are Becoming Software Systems12 minStep 05 · Production AI
Production AI
Inference, observability, cost control, and infrastructure — operating what you built.
Start withFrom Prototype to Production: AI System Design12 min
About the author
Vinod Kumar
Staff Software Engineer · ServiceNow · Delhi, IndiaI've spent 12+ years building distributed systems at scale — APIs, platform infrastructure, the kind of work where uptime isn't optional. Now I spend most of my time making agents, LLMs, and retrieval systems actually work inside production, not just in demos.
NeuralSys is where I write it all down — the kind of clear, honest engineering writing I wished existed when I started.
Newsletter
Get the notes before they go public.
I send new NeuralSys articles directly to subscribers — deep technical writing on AI agents, context engineering, RAG, and production AI architecture. No fluff. Unsubscribe anytime.
No spam. Unsubscribe anytime.