Accelerate with AI,
Turn it into value with judgment,
and grow it through operations.

I build both the product and the way I work around AI from the start. Agents handle generation and execution. I set the spec, verify the result, and decide what to publish. I measure the impact with data and keep improving in production.

See projects → About me →

9

Projects

9

Live

19+

yrs of development

Product work across
diverse industries

Working AI-Native

AI-native means building the structure of the work around AI, rather than adding it on top. Here is how that works on the product side and on the development side.

PRODUCTS BUILT ON AI

Remove the AI and the product stops

Products where inference sits at the centre of the experience.

On-device inference generates subtitles and explanations (Veloquo)
Models are routed by role, with Guardrails keeping the output within bounds (Relora)
A deterministic gate runs first; the LLM only classifies the input and writes the prose (MiraProof)

DEVELOPMENT RUN BY AGENTS

AI generates, people verify

I write the spec, let agents generate, verify the result, and feed what I learn back into the spec.

Ingest, translate, classify, and review articles automatically every day
Two models review each entry, and it publishes automatically only when they agree
A person reviews deviations and exceptions rather than every single change

The lifecycle of
AI products

I grow products by running this cycle, again and again.

01

Discover & Strategy

Frame user problems and business goals, then decide up front what AI does and does not touch.

02

Design & Data

Shape the experience, pick models, and build a reliable data foundation.

03

Build & Integrate

Integrate LLM/ML into the product and use guardrails to keep its output within bounds.

04

Evaluate

Keep quality stable with evaluation sets and regression tests, then verify value with data.

05

Operate & Improve

Decide what can ship automatically and what needs a human to stop it, with a takedown procedure ready before launch.

Featured Projects

See all →

iOS / Web

SmileProgress

An iOS app that aligns and anonymises orthodontic photos on-device for personal logging. Optional anonymous Before / After sharing, browsable on a web gallery.

SwiftUICoreMLAVFoundationStoreKit 2AWS LambdaNext.js 16

Read about SmileProgress →

Web / Tools

SiteKensa

Enter a domain name and get its email authentication (SPF / DKIM / DMARC), DNS, SSL, and search-engine visibility diagnosed — with fix steps matched to the services in use and record values you can paste as they are.

Next.js 16TypeScriptAWS Lambda

Read about SiteKensa →

iOS / iPadOS

Veloquo

Turn English videos on iPhone or iPad into a dual-subtitle learning experience. Subtitle generation, translation, and explanations all run on-device.

SwiftUISpeechTranslation

Read about Veloquo →

Latest Articles

Read all →

Source	Title	Lang	Published
Blog	Can't Sleep Hungry? What to Eat Before Bed, and How Much	JPEN	Aug 24, 2026
Blog	SPF Void Lookup Limit: When a Third Lookup Triggers PermError	JP	Aug 23, 2026
Blog	Fixing SPF Record Syntax Errors — How to Identify the Problem and Replace ptr	JP	Aug 22, 2026
Blog	SPF Record all Qualifier Explained: ~all vs -all — Which Should You Use	JP	Aug 19, 2026
Blog	"DKIM May Not Be Configured" — 3 Steps to Verify It	JP	Aug 17, 2026