Selected builds / 2026

Systems built from realoperating constraints.

Three examples of Benjamin's approach: identify the costly bottleneck, design the smallest reliable system, ship it into live operations, then measure what changed.

Ecommerce / Product / Shopify

Roadshow POS

8 years of checkout friction became a working tablet product in 3 days.

Problem

Public-site checkout was slow and venue Wi-Fi was unreliable. A clipboard-to-cashier workaround created queues and SKU errors. A custom overlay had remained too expensive to justify.

System

Benjamin built a tablet-first overlay on the Shopify Admin API with a custom product picker, payment-link generation, draft-order creation and real-time warehouse inventory.

Result

The system entered live use at the next show, removed SKU drift and reduced roadshow closing time by more than 50%.

What it shows

Technical product ownership grounded in years of operational context, with AI reducing build cost rather than replacing business judgment.

Read the build account on LinkedIn

AI operations / Data / Automation

Operator OS

A fragmented reporting routine became an automated operating briefing.

Problem

Critical numbers lived across Shopify, Klaviyo, Meta and Google. Manual weekly assembly consumed Benjamin's best thinking time and still produced a stale picture.

System

A deterministic early-morning pipeline pulls source data, checks anomalies against rolling baselines and uses an AI layer to produce a structured 400-word briefing.

Result

2 hours of assembly became 5 minutes of analysis. The system now runs across two ventures as part of a broader agent-supported operating stack.

What it shows

Disciplined separation between deterministic business logic and AI judgment, designed around reliability, latency and decision quality.

Applied AI / Measurement / AEO

AI Visibility Snapshot

A scoring engine that measures whether AI models actually recommend a brand.

Problem

Brands can see their Google rankings but have no idea what ChatGPT, Claude, Perplexity, Gemini or Google AI Overviews say when a buyer asks for a recommendation. No baseline means no way to prove any intervention works.

System

Benjamin built a scoring pipeline that runs a brand's real buyer queries across five AI models and grades the answers on seven dimensions, producing a repeatable visibility score.

Result

Across the Singapore DTC and retail brands scored so far, the same gap repeats: strong on their own brand name, near-invisible on the category-education queries buyers open with. Most land in the lower-mid quartile before any work. The engine now powers the assessment work at Firstpress.ai.

What it shows

Building the measurement instrument before selling the intervention.

Explore Firstpress.ai