Blogs

Notes from production.

Practical notes on the parts of AI engineering that are harder than the demo: OCR and extraction accuracy you can actually defend, integrating machine learning with the systems you already run, and what it takes to keep a model useful six months after launch.

MCP
Building an MCP server for internal tools: what we got wrong first
Our first server mirrored our REST API and the model used it badly. Five changes that made the difference.
20 Aug 20266 min
Read
MCP
What is MCP, and when is it worth building a server?
Model Context Protocol in plain terms: what problem it solves, what it does not, and how to tell whether your team needs one.
16 Jul 20266 min
Read
Shipping AI
Human-in-the-loop design: deciding when to ask a person
Route too much to review and you have rebuilt the manual process. Route too little and people stop trusting the output. Where the line goes.
18 Jun 20265 min
Read
LLM engineering
Fine-tuning, RAG or prompting: choosing without wasting a quarter
Three techniques that solve different problems and get picked for the wrong reasons. A decision order that avoids the expensive mistake.
21 May 20266 min
Read
Web & mobile
Choosing between React Native and native for a small team
Cross-platform is not a compromise, and native is not a luxury. The four questions that actually decide it.
30 Apr 20265 min
Read
Computer vision
Edge or cloud for video analytics: what it actually costs
The cloud version looks cheaper until you price the bandwidth. A worked comparison for a small multi-site deployment.
9 Apr 20266 min
Read
Document AI
Parsing Indian invoices: GSTIN, HSN and the formats that break extraction
Extraction tools trained on US and European paperwork miss the fields that matter here. What is different, and what to validate.
19 Mar 20266 min
Read
Shipping AI
What actually breaks in a production ML system
It is almost never the model weights. Four failure modes we keep meeting, and the monitoring that catches each one.
26 Feb 20265 min
Read
LLM engineering
How to build an LLM eval set when you have no labelled data
You cannot improve what you cannot measure, and nobody hands you a labelled set. How to build a useful one in about a day.
5 Feb 20266 min
Read
Web & mobile
Prerendering a React SPA for SEO without switching to Next.js
A Vite SPA serves crawlers an empty div. You can fix that in about a hundred lines without adopting a framework or a server.
15 Jan 20266 min
Read
Document AI
Field-level or document-level? How to measure extraction accuracy
The same pipeline can be 97% accurate or 68% accurate depending on how you count. Which denominator to use, and when each one lies.
16 Dec 20255 min
Read
Computer vision
Running computer vision on CCTV cameras you didn’t choose
Nobody replaces sixty cameras to try analytics. What it takes to get useful results from mixed, badly angled, years-old hardware.
27 Nov 20256 min
Read
Shipping AI
How to price an AI project when you can’t promise accuracy
Fixed-price assumes a known scope. Machine learning does not give you one up front. A structure that protects both sides without pretending.
6 Nov 20255 min
Read
LLM engineering
Why your RAG demo works and your RAG product doesn’t
A retrieval demo on twenty documents proves almost nothing about the same system on twenty thousand. The four things that break in between.
14 Oct 20256 min
Read
Document AI
Why your invoice OCR works in testing and fails in production
The test folder is clean exports. The real intake is phone photographs. What changes between the two, and how to find out before a client does.
18 Sept 20255 min
Read

Have a problem
worth solving?

Tell us what you're building. We'll help you work out what's actually possible, and say so if we're not the right people for it.

Start a conversation See the work

Notes from production.

Building an MCP server for internal tools: what we got wrong first

What is MCP, and when is it worth building a server?

Human-in-the-loop design: deciding when to ask a person

Fine-tuning, RAG or prompting: choosing without wasting a quarter

Choosing between React Native and native for a small team

Edge or cloud for video analytics: what it actually costs

Parsing Indian invoices: GSTIN, HSN and the formats that break extraction

What actually breaks in a production ML system

How to build an LLM eval set when you have no labelled data

Prerendering a React SPA for SEO without switching to Next.js

Field-level or document-level? How to measure extraction accuracy

Running computer vision on CCTV cameras you didn’t choose

How to price an AI project when you can’t promise accuracy

Why your RAG demo works and your RAG product doesn’t

Why your invoice OCR works in testing and fails in production

Have a problemworth solving?

Have a problem
worth solving?