#ai-engineering articles

Open Knowledge Format (OKF)

Imagine asking your company's AI assistant: “What is our approved definition of net revenue, which payment tables should I use, and did the refund policy change last quarter?” A conventional RAG ass

2 sanjaysorathiya.hashnode.dev

Ignacio Di Rocco2d ago · 6 min read

A Small Pipeline, With Explicit Boundaries: Architecture & Implementation

AI Waiter started with a focused goal: generate synthetic restaurant customer–waiter interactions that could be inspected, labeled, and evaluated—not just plausible-looking text. I wanted to experimen

idirocco.hashnode.dev

CapeStart2d ago · 10 min read

How Forward-Deployed Engineering Is Rewriting Enterprise AI Delivery

Forward-Deployed Engineering: An Overview Enterprise AI has reached a familiar and costly plateau. Models are more capable than ever. Cloud platforms and open-source tooling have removed many barriers

cstechnologyblog.hashnode.dev

RONI DAS6d ago · 5 min read

8 ms in the Notebook, 4 Seconds in Production: Serving a Model as an API (AI Engineering in 30 Days, Day 4)

A team ships a recommendation model. In the notebook, model.predict(x) returns in 8 milliseconds. They wrap it in a quick Flask app, put it behind a URL, and it works for the first ten users. Then rea

systemdesignacademy.hashnode.dev

RONI DAS6d ago · 12 min read

We Paid Double for the Same AI Model. The Reply Came Back 7% Sooner.

At 09:48 on a Sunday morning we started sending the same question to the same AI model, over and over. Half of the requests were marked "priority", which costs twice as much. Seven minutes and 150 req

systemdesignacademy.hashnode.dev

RONI DASOct 24 · 7 min read

Six Minutes or Eight Seconds: Packaging a Model So It Runs Anywhere (AI Engineering in 30 Days, Day 3)

You change one line of your model server's code and rebuild the container. It takes six minutes. A colleague moves one line in the Dockerfile, makes the same change, and the rebuild takes about eight

systemdesignacademy.hashnode.dev

RONI DASOct 23 · 7 min read

180 maths questions, 9 right answers, and an AI that said it was 95 to 100% sure almost every time

Many AI products are built on a large language model (LLM), the kind of model behind ChatGPT. Many of them have a step that looks like this: the model answers, then the model is asked "how sure are yo

systemdesignacademy.hashnode.dev

Harshil Kumar PindikuruOct 22 · 24 min read

How the Chef Thinks: Orchestration & Design Patterns in Agentic AI

Imagine a chef who knows hundreds of recipes, understands every ingredient, and has access to a fully equipped kitchen. The chef can choose ingredients, operate equipment, and prepare individual dishe

harshils.hashnode.dev

Ignacio Di RoccoOct 19 · 8 min read

A Passing Eval Doesn't Mean Your AI Feature Is Reliable

When building an AI feature, it is tempting to treat a good evaluation score as the finish line. The pipeline runs, the examples look reasonable, the judge scores them highly. Time to ship? Not so fas

idirocco.hashnode.dev

Ignacio Di RoccoOct 15 · 14 min read

My LLM Judge Went From 70% to 96% Agreement With Humans. Here’s What I Learned.

An engineering diary about evaluating an AI waiter, debugging an LLM judge, and discovering that a better rubric can matter more than a cleverer model. When I started building the evaluation pipeline

idirocco.hashnode.dev

Search Hashnode

Trending tags this week

Open Knowledge Format (OKF)

A Small Pipeline, With Explicit Boundaries: Architecture & Implementation

How Forward-Deployed Engineering Is Rewriting Enterprise AI Delivery

8 ms in the Notebook, 4 Seconds in Production: Serving a Model as an API (AI Engineering in 30 Days, Day 4)

We Paid Double for the Same AI Model. The Reply Came Back 7% Sooner.

Six Minutes or Eight Seconds: Packaging a Model So It Runs Anywhere (AI Engineering in 30 Days, Day 3)

180 maths questions, 9 right answers, and an AI that said it was 95 to 100% sure almost every time

How the Chef Thinks: Orchestration & Design Patterns in Agentic AI

A Passing Eval Doesn't Mean Your AI Feature Is Reliable

My LLM Judge Went From 70% to 96% Agreement With Humans. Here’s What I Learned.