NJ
Software engineerBelfast, Northern Ireland

Nathan Joel.

I’m a Forward Deployed Engineer. I work with retail customers, build software, and help turn the problems we find into a better product.

Outside work, I spend time on photography and small projects for my own curiosity. This is where I keep a bit of both.

More about my work
A river running through a green mountain valley beneath snow-covered peaks in Gimmelwald.
GimmelwaldMore photographs ↗

How I work

My work sits between retail customers and an AI-driven content product. I’m on the customer calls, mapping out their systems and requirements, and I also write the code that connects those systems.

That includes Python data pipelines, full-stack features, and internal tools. When the same manual task keeps appearing across accounts, I take it through design, development and testing with the people who will use it.

Based in Belfast. Working with retailers across the UK and North America.

Work

Three parts of my day-to-day work.
From the first customer conversation to a running system.

Customer delivery

Retail integrations & customer delivery

Technical discovery, systems mapping, and marketplace go-live across a portfolio of 15+ brands.

Read more

In the room for six to ten customer calls a week — from first technical scoping to data live on their marketplace.

I sit alongside Customer Success on most of their calls as the technical half of the conversation. A lot of that is mapping each retailer’s world before we touch it — how their PIM is set up, where product data actually lives, which channels and marketplaces they sell through — and then working out the shortest honest path to getting our data live in those systems.

The rest is keeping a portfolio of 15+ brands moving at once: standing up new accounts, unblocking stalled work, handling the edge cases that only appear at scale, and staying close enough to each account to catch the requirement nobody thought to write down.

  • Customer calls
  • PIM & systems discovery
  • Marketplace go-live
  • Technical support
  • CS partnership

Data & AI engineering

Data pipelines & AI content systems

Pipelines that turn inconsistent retailer data into useful inputs, plus the tools to run and review AI-generated content.

Read more

Build the data pipelines that feed the AI — and ship features into the content system itself.

Good product copy starts with good data, and retailer data is rarely clean — specs hidden behind dropdowns, the same attribute named five different ways, dropdown options that look like features but aren’t. I build the pipelines that pull all of that in, untangle it, translate it across languages, and hand the AI something it can actually work with.

On the generation side, the system writes titles, descriptions and metadata in stages — research, extraction, drafting, checking — so each step can be tuned without breaking the others. I ship full-stack features across it: the tools operators use to run and review jobs, and the quality checks that catch the failures you only see at volume.

  • Python
  • TypeScript
  • Enrichment pipelines
  • Data standardisation
  • Applied AI

Product engineering

Turning recurring tasks into product features

Taking recurring manual work from a design proposal through prototyping, full-stack development, and user testing to release.

Read more

Take a recurring bottleneck through the full product pipeline — from proposal to shipped, tested feature.

The most valuable thing I do is notice when the same manual work is showing up on every other account and treat it as one product problem instead of twenty client problems. Most recently that meant identifying the bottleneck, writing the design proposal that became the product direction — and then carrying it the whole way rather than handing it off: whiteboarding flows in FigJam, sketching and prototyping the interface in Pencil, and building it against a design spec agreed with the team.

The build itself ran on a disciplined AI-assisted workflow — research, plan, then implement with Claude Code — with well-shaped pieces delegated to teammates so the work parallelised instead of bottlenecking on me. Before release I put it in front of the CS team for user testing, then prioritised and actioned their feedback into the shipping version. Proposal to production, one continuous thread.

  • Design proposals
  • FigJam & Pencil
  • Claude Code
  • Delegation
  • User testing with CS
  • Feedback triage

Open tabs

Two projects I’m learning through.
Still a work in progress.

Research in progress

Recognising voices

A Shazam-style concept for voice actor recognition. How do you turn a voice into a fingerprint, search millions of those fingerprints, and isolate one speaker from a crowded track? Some of it is prototyped. Most of it is still notebooks and questions.

Voice fingerprints · similarity search · speaker isolation

Photographs

Places I’ve been, on foot and from above.
All photographs are my own.

Aerial view of the church on the island in Lake Bled, surrounded by turquoise water and distant mountains.
Bled
Himeji Castle framed by pink blossom, green foliage and a bright turquoise sky.
Himeji Castle
An aerial view of Bo-Kaap’s hillside homes painted vivid pink, turquoise, orange, yellow and green.
Bo-Kaap